Heng Huang, Ph.D. - US grants

This is funding to support a Doctoral Consortium of approximately 12 promising doctoral students from U.S. institutions of higher learning along with 8 distinguished research faculty, to be held in conjunction with the 1st International Conference on PErvasive Technologies Related to Assistive Environments (PETRA08), which will take place July 15-19 in Athens, Greece. The goal of the PETRA08 conference is to bring together experts from diverse domains to address an important social and healthcare issue, namely that as the world's population ages there is an urgent need to develop solutions for in-home care of the elderly, as well as of people with Alzheimer's, Parkinson's, and other disabilities or traumas. PETRA08 will provide a unique venue that focuses on combining wireless computing, sensors, and other pervasive computing technologies to assistive environments; unlike traditional computer science conferences, it will create a channel for applying basic CS principles to the service of millions of humans in need. More information about PETRA08 may be found at www.petrae.org; the organizers hope the conference will become an annual event. The goals of the Doctoral Consortium are to increase the exposure and visibility of the participants' work within the community, to help establish a sense of community among this next generation of researchers, and to help foster their research efforts by providing substantive feedback and guidance from a group of senior researchers in a supportive and interactive environment. Student participants in the Doctoral Consortium will make formal presentations of their work and will receive feedback from a faculty panel; the feedback is geared to helping students understand and articulate how their work is positioned relative to other research, whether their topics are adequately focused for thesis research projects, whether their methods are correctly chosen and applied, and whether their results are appropriately analyzed and presented. Doctoral Consortium attendees will have short papers on their work included in the Conference Proceedings, and a summary report on the event will be posted on the conference website.

Broader Impacts: The PETRA08 Doctoral Consortium will bring together some of the best students, researchers and practitioners in relevant fields, and will thereby afford the younger participants a unique opportunity to gain wider exposure for their innovative ideas while also receiving reinforcement for the importance and value of conducting research with societal impact. The workshop will also allow the junior participants to create a social network both among themselves and with senior colleagues. Since the conference is expected to host a diverse group along several dimensions (such as nationality, scientific discipline, and research specialization), participants? horizons will be broadened and new collaborations will emerge, to the future benefit of the field.

Collaborative Research: Matrix-Model Machine Learning: unifying machine learning and scientific computing

To analyze the ever-growing massive quantities of data for pattern recognition and knowledge discovery, effective machine learning models and efficient computational algorithms are essential tools. The goal of this research is to establish a theoretical foundation for solve challenging machine learning problems utilizing matrix/tensor computational methodologies, leveraging over the success of scientific computing over recent decades - including well-developed algorithms and mature, freely-available software.

This research begins with a critical connection between machine learning and scientific computing: an effective global solution to K-means clustering algorithm is provided by the principal component analysis which is based on singular value decomposition (SVD). This fundamental relationship will be systematically extended to matrices, tensors and multi-relational data, to deal with increasingly higher dimensions, multiple indexes and data types. The key goal of this research is to establish that well-known scientific computing techniques such as SVD, matrix and tensor decompositions can be directly utilized for pattern discovery, and further develop these computational methodologies for semi-supervised learning, clustering and classification. The focus will be on multi-index data (tensors, such as a sequence of weather maps or a sequence of traffics over a network) and multi-relational data (multiple pairwise relations, such as protein domains ? proteins ? pathways or words ? documents ? authors). Applications in genomics, text mining, and computer vision will be investigated.

Nonnegative matrix factorization (NMF) factorizes an input nonnegative matrix into two nonnegative matrices of lower rank. It was recently discovered that NMF has unique ability to solve challenging data mining and machine learning problems. The advantage of NMF over existing unsupervised learning methods are (1) NMF can model widely varying data distributions, (2) NMF performs both hard and soft clustering simultaneously. (3) Many other data mining problems such as semi-supervised clustering problems can be reformulated as NMF problem. Building upon these foundations, the investigators propose to establish a NMF-based comprehensive framework for data mining: (a) Provide deeper understanding of NMF's clustering capability;
(b) Extend data mining capability of NMF for solving various data mining and machine learning problems; (c) Develop fast numerical algorithms which incorporate the state-of-the-art developments from numerical optimization for various matrix factorization models; (d) Develop novel and rigorous proof strategies to prove the correctness and convergence properties of the numerical algorithms; (e) Apply and evaluate these new algorithms in real-world applications.


The proposed work creates a new paradigm of analyzing vast amount of data and discovering new knowledge from the data by transforming established matrix computational methodologies. This new technology can automatically group news articles into meaningful categories, discover protein modules in protein networks, extract weather patterns in climate data, segment pictures into distinct objects, detect communities on the Web, and enable many other scientific discoveries and new technologies creation. On a fundamental level, the proposed work establishes that a simple matrix factorization in fact solves challenging data mining problems. This research reinforces the importance of mathematics in today's data centric world and encourages students to learn mathematics.

EAGER: Collaborative Research: Cross-domain Knowledge Transformation via Matrix Decompositions

Traditional data mining algorithms discover knowledge in new domains starting from the scratch, ignoring knowledge learned in other domains. Knowledge transformation is a transformative paradigm that utilizes previously acquired knowledge in other domains to guide knowledge discovery process in a new domain and is especially useful for large data sets. In particular, utilizing applicable knowledge in other domains helps to stabilize the unsupervised learning and generate results that we may have preliminary understanding.

The goal of this project is to design and develop cross-domain knowledge transformation mechanisms for knowledge discovery. The transformation mechanisms are based on matrix decompositions where the knowledge been transferred are represented directly and explicitly ? making them easy to comprehend and be utilized in practice. The proposed mechanisms provide a versatile knowledge transformation framework with solid theoretical foundation and enable a new paradigm of unsupervised learning with domain knowledge.

The usefulness of these knowledge transformation mechanisms/systems will be demonstrated for effective information retrieval, consumer recommender systems, and product/online opinion sentiment analysis. The versatility of this transformative metholody will be verified across many domains.

This is funding to support a Doctoral Consortium of approximately 12 promising doctoral students from U.S. institutions of higher learning along with 3 distinguished research faculty, to be held in conjunction with the Second International Conference on PErvasive Technologies Related to Assistive Environments (PETRA 2009), which will take place July 9-13 in Corfu, Greece. The goal of the PETRA 2009 conference is to bring together experts from diverse domains to address an important social and healthcare issue, namely that as the world's population ages there is an urgent need to develop solutions for in-home care of the elderly, as well as of people with Alzheimer's, Parkinson's, and other disabilities or traumas. PETRA 2009 will provide a unique venue that focuses on combining wireless computing, sensors, and other pervasive computing technologies into assistive environments. The conference aims to span the continuum from data involving genetic and brain imaging to behavioral patterns; it will encompass and merge security and privacy issues with monitoring for both physical and digital safety in assistive environments including the home, work place, hospital, rehabilitation / nursing home, etc. Thus, PETRA 2009 will create a channel for applying basic CS principles to the service of millions of humans in need. More information about PETRA 2009 may be found at www.petrae.org. The goals of the Doctoral Consortium are to increase the exposure and visibility of the participants' work within the community, to help establish a sense of community among this next generation of researchers, and to help foster their research efforts by providing substantive feedback and guidance in a supportive and interactive environment from a group of senior researchers. Student participants in the Doctoral Consortium will make formal presentations of their work and will receive feedback from a faculty panel; the feedback is geared to helping students understand and articulate how their work is positioned relative to other research, whether their topics are adequately focused for thesis research projects, whether their methods are correctly chosen and applied, and whether their results are appropriately analyzed and presented. Doctoral Consortium attendees will have short papers on their work included in the Conference Proceedings, and a summary report on the event will be posted on the conference website.

Broader Impacts: The PETRA 2009 Doctoral Consortium will bring together some of the best students, researchers and practitioners in relevant fields, and will thereby afford the younger participants a unique opportunity to gain wider exposure for their innovative ideas while also receiving reinforcement for the importance and value of conducting research with societal impact. The workshop will also allow the junior participants to create a social network both among themselves and with senior colleagues. Since the conference is expected to host a diverse group along several dimensions (such as nationality, scientific discipline, and research specialization), participants' horizons will be broadened and new collaborations will emerge, to the future benefit of the field. The organizing committee will make a concerted effort to attract minority and disabled participants to the event.

Proposal #: CNS 09-23494
PI(s): Makedon, Fillia S.; Athitsos, Vassilis; Huang, Heng; Le, Zhengyi; Popa, Dan O.
Institution: University of Texas - Arlington
Title: MRI/Dev.: Next Generation Multimodal Data Management Human-Sensing Instrument for
Trustworthy Research Collaboration and Quality of Life Improvement
Project Proposed:
This project, developing an instrument that serves as an interactive personal care and human activity monitoring center, aims to keep a person with high quality life and safe at home as long as possible. The instrument enables privacy-preserving and secure data sharing through wireless connection with remote users in an assistive living environment. Providing mental and emotional support, the zooscopion (zScope) can connect devices, humans, objects, and the environment. It can connect to other assistive living projects, making them interoperable and can deliver a Digital Library of sanitized research data and cases with high educational and training value. zScope combines and correlates many types of data and extracts events of interest that indicate changes, risks, etc. It can analyze facial expressions to detect pain, environmental data, house data (such as door opening, telephone sounds, vacuum cleaner, etc.), human performance metrics (e.g., hand strength), both in continuous and discrete format. Data are modeled and assembled in meaningful ways to predict and prevent physical and digital problems (e.g., respectively, falls and intrusions). Privacy and security are being made part of the data modeling at the design phase. The instrument will take sensor data, human body measurements, camera data when requested, known pattern of behavior from other cases, brain scans, and clinical information, aiming to provide high resolution displays of longitudinal as well as episodic events. It outputs a visual interactive display of patterns and significant human behavioral 'events' valuable in assistive environments, setting where to use non-invasive monitoring technologies, helping recognize 'behavioral biomarkers' that will be connected to other types of health indicators that may come from brain imaging, genetic analysis, clinical results, or psychological evaluations. It will work with the next generation of data that include behavioral, clinical, body motion, etc., and have low latency tracking.
Broader Impacts:
This work enables human-centric type of experiments and provides novel new ways of interaction, visualization, and secure collaboration. Developing 'smarter' living environments for the aged opens new ways to education with immersive compelling projects that provide a better understanding of the role of science and engineering when combining health data (genomic information) to behavior, predict trends, and provide indicators of how medication and clinical assessments connect to longitudinal behavior. Long term goals include behavioral markers for assessing the confluence of environment, drugs, and human psychology. The instrument is also expected to respond to queries regarding emerging needs for new analysis of collected information. It includes training and educational modules with search and browsing tools and a recommender facility to support decision making and use stored strategies. Moreover, utilizing existing outreach programs, the project will support local minority students and high school students A new generation of scientists that can work together across domain silos towards human centered goals might be in the making!

New Theoretical Foundations of Tensor Applications: Clustering, Error Analysis, Global Convergence, and Robust Formulations

Tensor decompositions become increasingly important in analyzing high-dimensional multi-index data. However, applications of tensor decompositions are so far restricted: (1) they are mainly used for data compression ? critically important tasks such as data clustering have not been addressed. (2) No bounds on reconstruction error exist ? the compression parameters are determined on a trial-and-error basis. (3) As solutions to non-convex optimizations, tensor decompositions are not unique. This could severely affect the reliability of tensor analysis. (4) Tensor decompositions are obtained via minimizing the sum of squared errors, thus are prone to noise or outliers in the data. A robust formulation of decomposition is highly desirable for applications with large noises. In this proposal, we investigate these new fundamental aspects of tensor applications:
(1) Investigate the clustering capabilities of tensor decompositions, in addition to the established theoretical results on clustering;
(2) Provide comprehensive error analysis of tensor decompositions and derive lower and upper error bounds;
(3) Investigate conditions for global convergence for tensor decompositions and investigate good initializations for the cases where global convergence fails.
(4) Develop robust formulations for tensor decompositions.
In addition, we will develop user-friendly software toolbox that contains the resulting algorithms and make it available to the public. We will also educate graduate and undergraduate students with fundamentals in matrix and tensor computations. We will present tutorials and organize workshops on this new direction.

This award provides NSF support to sponsor students and faculty mentors from US institutions to participate in a doctoral symposium held during the First Workshop on Privacy and Security in Pervasive e-Health and Assistive Environments (PSPAE?09 http://www.petrae.org/docs/PSPAE_ann_f.pdf ) in Corfu, Greece, June 9-13, 2009. This workshop is in conjunction with the Second International Conference on Pervasive Technologies Related to Assistive Environments (PETRA?09 http://www.petrae.org).

The doctoral symposium shares the aims of the PSPAE workshop, seeking new directions in security and privacy research and education. The focus of the PSPAE workshop is a multidisciplinary discussion aimed towards developing new insights into privacy and security issues for pervasive technologies that will be deployed for the benefit of society. The scientific community is expected to benefit from discussions on the crossover between security and privacy and on the issues arising in complex application areas such as healthcare. The doctoral consortium seeks to open new horizons for graduate students who may pursue privacy and security research in healthcare and other pervasive assistive environments. Another objective is to create an international forum that can expose graduate students to the differences in privacy policies around the world, especially regarding health privacy and security.

The objective of this research is to develop methods and tools for a multimodal and multi-sensor assessment and rehabilitation game system called CPLAY for children with Cerebral Palsy (CP). CPLAY collects and processes multiple types of stimulation and performance data while a child is playing. Its core has a touch-screen programmable game that has various metrics to measure delay of response, score, stamina/duration, accuracy of motor/hand motion. Optional devices attached to extend CPLAY versions provide additional parallel measurements of level of concentration/participation/engagement that quantify rehabilitation activity. The approach is to model the process as a cyber-physical system (CPS) feedback loop whereby data collected from various physical 3D devices (including fNIR brain imaging) are processed into hierarchical events of low-to-high semantic meaning that impact/ adjust treatment decisions.

Intellectual Merit: The project will produce groundbreaking algorithms for event identification with a multi-level data to knowledge feedback loop approach. New machine learning, computer vision, data mining, multimodal data fusion, device integration and event-driven algorithms will lead towards a new type of cyber- physical rehabilitation science for neurological disorders. It will deliver fundamental advancements to engineering by showing how to integrate physical devices with a computationally quantitative platform for motor and cognitive skills assessment.

Broader Impacts: The project delivers a modular & expandable game system that has huge implications on the future of US healthcare and rehabilitation of chronic neurological disabilities. It brings hope to children with Cerebral Palsy via lower cost and remote rehabilitation alternatives. It brings new directions to human centered computing for intelligent decision-making that supplements evidence-based practices and addresses social and psychological isolation problems.

This is funding to support a doctoral consortium (workshop) of approximately 10 promising graduate students from U.S. institutions of higher learning along with 3 distinguished research faculty, to be held in conjunction with the Third International Conference on Pervasive Technologies Related to Assistive Environments (PETRA 2010), which will take place June 23-25, 2010, on the island of Samos, Greece (http://www.petrae.org). PETRA is the leading annual interdisciplinary conference on assistive technologies. Its brings together experts in healthcare, sensors,, wireless communications, smart devices, intelligent software, privacy and security, to address an important social and healthcare issue, namely that as the world's population ages there is an urgent need to develop solutions for in-home care of the elderly, as well as of people with Alzheimer's, Parkinson's, and other disabilities or traumas. PETRA provides a unique venue that focuses on combining wireless computing, sensors, and other pervasive computing technologies into assistive environments. While PETRA 2010 will continue to bridge the continuum from data collection and processing to semantic understanding of human behavior, it will also evolve and incorporate new exciting aspects such as how to connect the genotype with phenotype, or the clinical/biomedical features with behavioral patterns and how these impact each other and thus refine research. PETRA 2010 will also study vital issues in privacy and security in monitoring ambient intelligent environments with the goal of identifying and predicting risks, intrusions, unauthorized access to information, or information leaking. PETRA projects assume a human is at the center of "cyberphysical systems" where the digital world merges with the physical.

The goals of the Doctoral Consortium are to increase the exposure and visibility of the participants' work within the community, to help establish a sense of community among this next generation of researchers, and to help foster their research efforts by providing substantive feedback and guidance in a supportive and interactive environment from a group of senior researchers. Student participants in the Doctoral Consortium will be drawn from diverse communities including computer science, engineering, psychology, social science, neuroscience, human-computer interaction, cognitive science and communication. They will make formal presentations of their work and will receive feedback from a faculty panel; the feedback is geared to helping students understand and articulate how their work is positioned relative to other research, whether their topics are adequately focused for thesis research projects, whether their methods are correctly chosen and applied, and whether their results are appropriately analyzed and presented. The workshop faculty members will bring a wide spectrum of expertise, and provide student mentoring and coordination. Doctoral Consortium attendees will have short papers on their work included in the Conference Proceedings, and a summary report on the event will be posted on the conference website.

Broader Impacts: The PETRA 2010 Doctoral Consortium will bring together some of the best students, researchers and practitioners in relevant fields, and will thereby afford the younger participants a unique opportunity to gain wider exposure for their innovative ideas while also receiving reinforcement for the importance and value of conducting research with societal impact. The workshop will allow the junior participants to create a social network both among themselves and with senior colleagues. Since the conference is expected to host a diverse group along several dimensions (such as nationality, scientific discipline, and research specialization), participants' horizons will be broadened and new collaborations will emerge, to the future benefit of the field. The organizing committee will make a concerted effort to attract participants who are women, members of under-represented minorities, and persons with disabilities.

Today's massive generation of digital data is greatly outpacing the development of computational methods and tools and presents critical challenges for achieving the full transformative potential of these data. For example, recent advances in acquiring multi-modal brain imaging and genome-wide array data provide exciting new opportunities to study the influence of genetic variation on brain structure and function. Major computational challenges are, however, bottlenecks for comprehensive joint analysis of these data due to their unprecedented scale and complexity. This project will employ the new capabilities of large-scale data mining techniques in multi-view learning, multi-task learning, and robust classification to address critical challenges in systematically analyzing massive multi-modal genetic, imaging, and other biomarker data. Specifically, this project will: (1) develop new multi-view learning methods to detect task-relevant phenotypic biomarkers from large scale heterogeneous imaging and other biomarker data, (2) implement new sparse multi-task regression models to reveal the genetic basis of phenotypic biomarkers at multiple levels (e.g., SNP, haplotype, gene and/or pathway), (3) design novel robust classification methods via structural sparsity for outcome prediction using integrated genotypic and phenotypic data, and (4) package these new methods into a data mining toolkit and release it to the public.

The intellectual merits of this project derive not only from the development of novel data mining methods, but also from their application to imaging genetic studies. These methods are designed to take into account interrelated structures among multiple data modalities and offer systematic strategies to reveal structural imaging genetic associations. The proposed methods and tools are expected to impact neurological and psychological research and enable investigators to effectively test imaging genetics hypothesis and advance biomedical science and technology. In addition, the proposed data mining framework addresses generic critical needs of large-scale data analysis and integration and, therefore, will impact a large number of research areas where high-value knowledge and complex patterns can potentially be discovered from massive high-dimensional and heterogeneous data sets. This project will facilitate the development of novel educational tools to enhance several current courses at UT Arlington and IUPUI. Both universities are minority-serving institutions, and the PIs will engage the minority students and under-served populations in research activities to give them a better exposure to cutting-edge scientific research.

This is funding to support a doctoral consortium (workshop) of approximately 10 promising graduate students from U.S. institutions of higher learning along with 9 distinguished research faculty, to be held in conjunction with the Fourth International Conference on Pervasive Technologies Related to Assistive Environments (PETRA 2011), which will take place May 25-27, 2011, on the island of Crete, Greece (http://www.petrae.org). PETRA is the only annual conference that brings together theoreticians and practitioners from a wide variety of disciplines to focus on the application of pervasive technologies to assistive environments. What connects different disciplines in this highly interdisciplinary but topical conference is the ability to discover new alliances, share data and design new methods that have real-world social impact while also advancing each field scientifically. PETRA 2011 will address timely computational challenges including integration mechanisms of diverse sensor technologies, modeling heterogeneous data, synthesis and analysis of data streams, privacy and security of information, ranking, cleaning, storing and retrieving sensor data streams for pattern analysis and discovery, robust remote rehabilitation mechanisms, correlating brain imaging with behavioral imaging, handling intensive real time data in complex environments, and ways to interpret seemingly meaningless data in order to derive meaningful human behavioral patterns and/or to identify important "events" or changes. PETRA 2011 will also study vital issues in privacy and security in monitoring ambient intelligent environments with the goal of identifying and predicting risks, intrusions, unauthorized access to information, or information leaking. PETRA projects assume a human is at the center of "cyberphysical systems" where the digital world merges with the physical.

The goals of the Doctoral Consortium are to increase the exposure and visibility of the participants' work within the community, to help establish a sense of community among this next generation of researchers, and to help foster their research efforts by providing substantive feedback and guidance in a supportive and interactive environment from a group of senior researchers. Student participants in the Doctoral Consortium will be drawn from diverse communities including computer science, engineering, psychology, social science, neuroscience, human-computer interaction, cognitive science and communication. They will make formal presentations of their work and will receive feedback from a faculty panel; the feedback is geared to helping students understand and articulate how their work is positioned relative to other research, whether their topics are adequately focused for thesis research projects, whether their methods are correctly chosen and applied, and whether their results are appropriately analyzed and presented. The workshop faculty members will bring a wide spectrum of expertise, and provide student mentoring and coordination. Doctoral Consortium attendees will be asked to create and maintain "conference workbook" (webpage) where along with their work they include feedback on papers they attended and observations that they gathered (e.g., through meetings and discussions with conference speakers and other participants) that will apply to and impact their work. Short papers on the participants' work will be published in the Conference Proceedings (which are included in the ACM Digital Library), and a summary report on the event will be posted on the conference website.

Broader Impacts: The PETRA 2011 Doctoral Consortium will bring together some of the best students, researchers and practitioners in relevant fields, and will thereby afford the younger participants a unique opportunity to gain wider exposure for their innovative ideas while also receiving reinforcement for the importance and value of conducting research with societal impact. The workshop will allow the junior participants to create a social network both among themselves and with senior colleagues. The organizing committee will make a concerted effort to attract participants who are women, members of under-represented minorities, and persons with disabilities.

This is funding to support a doctoral consortium (workshop) of approximately 12 promising graduate students from U.S. institutions of higher learning along with 9 distinguished research faculty, to be held in conjunction with the Fifth International Conference on Pervasive Technologies Related to Assistive Environments (PETRA 2012), which will take place June 6-8 on the island of Crete, Greece (http://www.petrae.org). PETRA is the only annual conference that brings together theoreticians and practitioners from a wide variety of disciplines to focus on the application of pervasive technologies to assistive environments. What connects different disciplines in this highly interdisciplinary but topical conference is the ability to discover new alliances, share data and design new methods that have real-world social impact while also advancing each field scientifically. PETRA projects assume a human is at the center of "cyberphysical systems" where the digital world merges with the physical. PETRA 2012 will address timely computational challenges including integration mechanisms of diverse sensor technologies, modeling heterogeneous data, synthesis and analysis of data streams, privacy and security of information, ranking, cleaning, storing and retrieving sensor data streams for pattern analysis and discovery, robust remote rehabilitation mechanisms, correlating brain imaging with behavioral imaging, handling intensive real time data in complex environments, and ways to interpret seemingly meaningless data in order to derive meaningful human behavioral patterns and/or to identify important "events" or changes. PETRA 2012 will also incorporate new exciting aspects in improving the quality of human life and health, such as how clinical and biomedical indicators are connected with observed behavioral patterns over time to refine a particular treatment.

The goals of the Doctoral Consortium are to increase the exposure and visibility of the participants' work within the community, to help establish a sense of community among this next generation of researchers, and to help foster their research efforts by providing substantive feedback and guidance in a supportive and interactive environment from a group of senior researchers. Student participants in the Doctoral Consortium will be drawn from diverse communities including computer science, engineering, psychology, social science, neuroscience, human-computer interaction, cognitive science and communication. They will make formal presentations of their work and will receive feedback from a faculty panel; the feedback is geared to helping students understand and articulate how their work is positioned relative to other research, whether their topics are adequately focused for thesis research projects, whether their methods are correctly chosen and applied, and whether their results are appropriately analyzed and presented. The workshop faculty members will bring a wide spectrum of expertise, and provide student mentoring and coordination. Doctoral Consortium attendees will be asked to create and maintain "conference workbook" (webpage) where along with their work they include feedback on papers they attended and observations that they gathered (e.g., through meetings and discussions with conference speakers and other participants) that will apply to and impact their work. Short papers on the participants' work will be published in the Conference Proceedings (which are included in the ACM Digital Library), and a summary report on the event will be posted on the conference website.

Broader Impacts: The PETRA 2012 Doctoral Consortium will bring together some of the best students, researchers and practitioners in relevant fields, and will thereby afford the younger participants a unique opportunity to gain wider exposure for their innovative ideas while also receiving reinforcement for the importance and value of conducting research with societal impact. The workshop will allow the junior participants to create a social network both among themselves and with senior colleagues. The organizing committee will make a concerted effort to attract participants who are women, members of under-represented minorities, and persons with disabilities. To further assure participant diversity, the organizers will endeavor not to accept more than 2 students from any one institution.

There are many types of disabilities caused by arthritis, where systematic physical activity and physical therapy are essential to preventing further deterioration. The challenge here is that periods of severe pain are an obstacle to such activities. The relationship between pain and physical activity is very tricky; it typically involves all joints (e.g., hands, wrists, feet, knees) and is a major cause of reduced quality of life and disability. This is the case with Rheumatoid Arthritis (RA), a chronic systemic inflammatory disease, where preserving functional range of motion and enhancing cardiovascular health are primary goals of physical therapy. Persons with RA who exercise regularly show not only improvements in muscle strength and overall physical and health function, but also reduced mortality. But it has been shown that long-term engagement in exercise among patients with RA is poor and does not exceed 50% when patients are not supervised. This results in a huge cost to national health care and to national productivity.

The PI argues that computer technology (tools and interfaces) can help the RA patient remain motivated and maintain movement. In this exploratory project her team will develop some of the components for motion capture and data integration that will eventually be integrated into a system called RPLAY which will enable patients with RA to perform physical therapy at home while the system accurately monitors their joint motions, motor performance and other physiological indicators. Specific objectives include: development of preliminary tools for human-centric adaptive remote monitoring, including facial, hand, arm, and body motions, and range of motion; development of prototype interactive games to encourage patient rehabilitation, and of a prototype user interface through which the patient and therapist can communicate via avatars to promote physical activity; and development of tools for validating the physical therapy monitoring and assessment results and their integration into a visualization and report system.

Broader Impacts: This project addresses issues of human joint pain and physical inactivity, which adversely affect the everyday lives of millions of people. It will lay the foundations for integrating disparate human data of persons in similar conditions, which will advance and extend the relationship between engineering innovation and computational analysis. The project will provide a training experience for computer science PhD students in modeling human motion in innovative ways. The project will also lead to an interdisciplinary course for physical therapy and computer science students on improving health delivery and rehabilitation practices.

Proposal #: 13-38118 Collaborative Proposal #: 13-37866
PI(s): Makedon, Fillia S PI(s): Betke, Margrit
Athitsos, Vassilis; Gatchel, Robert J; Huang, Heng; Romero-Ortega, Mario I
Institution: University of Texas-Arlington INstitution: Boston Univeristy
Title: MRI/Dev :Collab Dev. of iRehab, an Intelligent Closed-loop Instrument for Adaptive Rehabilitation
Project Proposed:
This project, developing of an instrument referred to as iRehab, aims to enable personalized rehabilitation therapy for individuals suffering from brain injury, motor disabilities, cognitive impairments, and/or psychosocial symptoms. The instrument, a modular rehabilitation device, in its simplest form consists of a computer, a camera, and adaptive software for assessment and training of cognitive functions. In its final, most complex form, the instrument will integrate data from a 4-degree-of-freedom robotic-arm with gimbals and torque sensing, a Kinect sensor, multiple cameras, an eye-tracking device, a touch screen, a microphone, and an fNIRS brain imaging sensor.
The instrument will be developed in two phases. In the first phase, the investigators develop a Barrett robot arm. In the second phase, the instrument will extend to a Kinect sensor, multiple cameras, an eye-tracking device, and related low-cost components, along with the assessment software for assessing motor function and cognitive, emotional, and personality functioning.
iRehab consists integrates multidisciplinary methodologies and sensors to assess and assist the cognitive and physical rehabilitation of persons affected by various impairments. This work highly interdisciplinary work follows a cyber-physical approach. It provides new research opportunities across the fields of human-centered computing, computer vision, assistive technology, robotics, machine learning, and neuroimaging. This work advances research in human brain activity mapping, personalized medicine, and big data.
Broader Impacts:
The proposed instrument exhibits potential for large broader impact as it directly contributes to future healthcare and human wellbeing improving accessibility to affordable rehabilitation for a broad range of patients. The instrument is likely to accelerate the recovery of a large spectrum of injuries and diseases including those causing motor, neurological, and cognitive disorders. An education plan includes course development, internships, workshops and tutorials, and an on-line resource center. In addition to many educational impacts, impact will be felt on the fundamental research in the areas addressed.

The increasingly large amounts of Electronic Medical Record (EMR) data offer unprecedented opportunities for EMR data mining to enhance health care experiences for personalized intervention, improve different diseases risk stratifications, and facilitate understanding about disease and appropriate treatment. To solve the key and challenging problems in mining such large-scale heterogeneous EMRs, the investigators aim to develop: (i) new computational tools to automate the EMRs processing, including new techniques for filling in missing values using a new robust rank-k matrix completion method; (ii) annotation of unstructured free-text EMRs using multi-label multi-instance learning; (iii) a new sparse multi-view learning model to integrate heterogeneous EMRs to predict the readmission risk of Heart Failure (HF) patients and to support personalized intervention; (iv) novel methods for identifying the longitudinal patterns using high-order multi-task learning; (v) a nonparametric Bayesian model for predicting the event time outcomes of the HF patients readmission.

The sparse multi-view feature learning and robust multi-task longitudinal pattern finding algorithms have a broad range of applications beyond EMR data mining. Free dissemination of source implementations of the algorithms enable other researchers to further develop and apply the resulting techniques. In particular, the methods and tools are expected to impact other EMR and public health research. This project offers enhanced opportunities for research-based advanced training of students (including members of minorities and under-served populations) and integration of research results into curricula at the University of Texas at Arlington, the University of Texas Southwestern Medical Center at Dallas, and Southern Methodist University. For further information see the web site at: http://ranger.uta.edu/~heng/NSF-III-1302675.html

This is funding to support a doctoral consortium (workshop) of approximately 14 promising graduate students from U.S. institutions of higher learning, along with 6 distinguished research faculty (4 of whom are women), to be held in conjunction with the Sixth International Conference on Pervasive Technologies Related to Assistive Environments (PETRA 2013), which will take place May 29-31 on the island of Rhodes, Greece. The PETRA Conference mission is to promote interdisciplinary research that can improve the quality of life and empower people with greater capabilities using pervasive ambient intelligent environments. This is the only annual conference that brings together theoreticians and practitioners from a wide variety of disciplines to focus on the application of pervasive technologies to assistive environments. Innovations to be presented during this year's conference include intelligent human sensing, robotic devices to enable persons with disabilities, algorithms for data fusion, new computer aided rehabilitation methods, therapy game development that personalizes a game to an individual's needs, gesture recognition tools, medication management, remote secure communications and data sharing, remote health monitoring, and many other enabling technologies. More information about the conference may be found online at http://www.petrae.org.

The goals of the Doctoral Consortium are to increase the exposure and visibility of the participants' work within the community, to help establish a sense of community among this next generation of researchers, and to help foster their research efforts by providing substantive feedback and guidance in a supportive and interactive environment from a group of senior researchers. Student participants in the Doctoral Consortium will be drawn from diverse communities including computer science, engineering, psychology, social science, neuroscience, human-computer interaction, cognitive science and communication. They will make formal 20-minute presentations of their work and will receive feedback from a faculty panel; the feedback is geared to helping students understand and articulate how their work is positioned relative to other research, whether their topics are adequately focused for thesis research projects, whether their methods are correctly chosen and applied, and whether their results are appropriately analyzed and presented. The workshop faculty members will bring a wide spectrum of expertise, and provide student mentoring and coordination. Doctoral Consortium attendees will be asked to create and maintain a "digital workbook" where along with their work they include feedback on papers they attended and observations that they gathered (e.g., through meetings and discussions with conference speakers and other participants) that will apply to and impact their work. Short papers on the participants' work will be published in the Conference Proceedings (which are included in the ACM Digital Library), and a summary report on the event will be posted on the conference website.

Broader Impacts: The PETRA 2013 Doctoral Consortium will bring together some of the best students, researchers and practitioners in relevant fields, and will thereby afford the younger participants a unique opportunity to gain wider exposure for their innovative ideas while also receiving reinforcement for the importance and value of conducting research with societal impact. The workshop will allow the junior participants to create a social network both among themselves and with senior colleagues. The organizing committee will make a concerted effort to attract participants who are women, members of under-represented minorities, and persons with disabilities. To further assure participant diversity, the organizers will endeavor not to accept more than 2 students from any one institution, and in addition the PI plans to ask her dean for separate funds to support 2-3 exceptional undergraduates to participate in the conference.

This project builds a novel privacy-preserving framework with both new algorithms and software tools to: 1) evaluate the effectiveness of current identifier-suppression techniques for Electronic Healthcare Record (EHR) data; 2) de-identify and anonymize EHR data to protect personal information without significantly reducing the utility of data for secondary data analysis. The proposed techniques eliminate the violation of privacy through re-identification, and facilitate the secondary usage, sharing, publishing and exchange of healthcare data without the risk of breaching protected health information (PHI). This new privacy-preserving framework injects the ICD-9-CM-aware constraint-based privacy-preserving techniques into EHRs to eliminate the threat of identifying an individual in the secondary use of research data. The proposed technique and development can be readily adapted to other types of healthcare databases in order to ensure privacy and prevent re-identification of published data. The project produces groundbreaking algorithms and tools for identifying privacy leakages and protecting personal privacy information in EHRs to improve healthcare data publishing. New privacy-preserving techniques developed in this project lead towards a new type of healthcare science for EHRs. The project also delivers fundamental advancements to engineering by showing how to integrate biomedical domain knowledge with a computationally advanced quantitative framework for preserving the privacy of published EHRs. HIPAA has established protocols and industry standards to protect the confidentiality of PHI. However, our results demonstrate that, even with regard to health data that meets HIPAA requirements, the risk of re-identification is not completely eliminated. By identifying the security vulnerabilities inherent in the HIPAA standards, our research develops a more rigorous security standard that greatly improves privacy protections by applying state-of-the-art algorithms.

The developed data privacy-preserving framework has significant implications for the future of US healthcare data publishing and related applications. Specifically, the transition from paper records to EHRs has accelerated significantly since the passage of the HITECH Act of 2009. The Act provides monetary incentives for the "meaningful use" of EHRs. As a result, the quality and quantity of healthcare databases has risen sharply, which has renewed the public's fear of a breach of privacy of their medical information. This research work is innovative and crucial not only for facilitating EHR data publishing, but also for enhancing the development and promotion of EHRs. At the educational front, this project facilitates the development of novel educational tools to construct entirely new courses and laboratory classes for healthcare, data privacy, data mining, and a wide range of applications. As a result, it enhances the current instructional methods for teaching data privacy and data mining, and has compelling biomedical and healthcare applications that can facilitate learning of computational algorithms. This project involves both undergraduate and graduate students in the three participating institutions. The PIs make a strong effort to engage minority graduate and undergraduate students in research activities in order to increase their exposure to cutting-edge research.

Large-scale in situ hybridization (ISH) screens are providing an abundance of data showing spatio-temporal patterns of gene expression that are valuable for understanding the mechanisms of gene regulation. Knowledge gained from analysis of Drosophila expression patterns is widely important, because a large number of genes involved in fruit fly development are commonly found in humans and other species. Thus, research efforts into the spatial and temporal characteristics of Drosophila gene expression images have been at the leading-edge of scientific investigations into the fundamental principles of different species development. Drosophila gene expression pattern images enable the integration of spatial expression patterns with other genomic datasets that link regulator with their downstream targets. This project addresses the computational challenges in analyzing Drosophila gene expression patterns by leveraging a new bioinformatics software system. It focuses on designing principled bioinformatics and computational biology algorithms and tools that will integrate multi-modal spatial patterns of gene expression for Drosophila embryos' developmental stage recognition and anatomical ontology term annotation, and will infer gene interaction networks to generate a more comprehensive picture of gene function and interaction. The bioinformatics methods resulting from the project activities are broadly applicable to a variety of fields such as biomedical science and engineering, systems biology, clinical pathology, oncology, and pharmaceutics. Novel tools to enhance courses and research experiences for diverse populations of students are planned to broaden participation in science.

This project investigates three challenging problems for studying the Drosophila embryo ISH Images via innovative bioinformatics algorithms: 1) the sparse multi-dimensional feature learning method to integrate the multimodal spatial gene expression patterns for annotating Drosophila ISH images, 2) the heterogeneous multi-task learning models using the high-order relational graph to jointly recognize the developmental stages and annotate anatomical ontology terms, 3) the embedded sparse representation algorithm to infer the gene interaction network. It is innovative to apply structured sparse learning, multi-task learning, and high-order relational graph models to Drosophila gene expression patterns analysis and holds great promise for scientific investigations into the fundamental principles of animal development. The algorithms and tools as outcomes of this research are expected to help knowledge discovery for applications in broader scientific and biological domains with massive high-dimensional and heterogeneous data sets. This project facilitates the development of novel educational tools to enhance several current courses at University of Texas at Arlington. The PIs engage minority students and under-served populations in research activities to provide opportunities for exposure to cutting-edge scientific research. For further information see the web site at: http://ranger.uta.edu/~heng/NSF-DBI-1356628.html

Accurate human motion tracking and activity recognition are important in supporting numerous areas of computer science and engineering research, ranging from user modeling and human robot interaction to graphics and animation. A good resource of annotated datasets and annotation tools is particularly important in research involving physical human limitations, persons with chronic disabilities, such as post-stroke, ALS, Rheumatoid Arthritis, Cerebral Palsy, and others. Currently, a comprehensive community CRI with human activity data and analysis tools is not in place. The development of new methods and algorithms that will improve and enable real-world motion tracking applications is hampered by an inherent difficulty: the lack of large sets of training and testing data. This planning activity brings together experts in computer vision, machine learning, data mining, user interface design, assistive environments, human robot interaction, databases research, big data involving real time human activity, therapists, clinicians, device makers and sensor developers in order to identify the specific human activity data that should be collected and processed when building a repository of automatically and accurately annotated video data of human motion.

Specifically, a repository of automatically and accurately annotated video data of human motion has the potential to impact the research of several disciplines using human motion tracking and recognition by significantly improving the ability to recognize behavioral markers for assessing the confluence of environment, drugs, human psychology and thus extend clinical and psychological areas. However, due to big data issues (volume, velocity, variety, veracity, arising from multisensory visual input of human behavior), the tasks of automating data collection and especially annotation are non-trivial. Traditionally, video data of human motion are collected and annotated manually. Due to the large amount of data generated by video sensors, this task is very tedious and error prone. For this reason, most existing datasets are very small and specialized to the problem for which they were collected. This project will gather input from the research user communities on how to best develop such a CRI repository of human activity. A workshop will be organized with invited experts who will help identify the specific human activity data that should be collected and processed. The input solicited during the workshop will help define the features and facilities of the proposed Smart-Move repository, a spatiotemporal annotated human motion repository by focusing on the following: (1) determining the specific activities to be captured and the methods of multisensory data collection; (2) tools for accurate real-time annotation and storage of the collected data; (3) search facilities to enable researchers to contribute and provide feedback; (4) event recognition and data summarization tools to enable researchers to analyze their own new big data automatically; (5) feasibility of the proposed data collection and analysis using a strategically chosen set of multi-stream data to keep the overall project cost low; (6) limitations when modeling heterogeneous types of human activity data collected by different sensors or of different sample frequencies.

? DESCRIPTION (provided by applicant): Mild cognitive impairment (MCI) is a prodromal stage of Alzheimer Disease (AD), and the existing studies have suggested that the individuals with amnestic MCI tend to progress to probable AD at a rate of approximately 10% to 15% per year. Early prediction of MCI patients with high risk of conversion to AD is of great importance for timely therapy and possible delay of the disease. Recent advances in acquiring multi-dimensional and longitudinal brain imaging and genome-wide array data provide exciting new opportunities to study the influence of genetic variation on brain structure and function. Integrating such multi-dimensional and longitudinal imaging genomic data holds great promise for a system biology of the brain to better understand the complex neurobiological mechanism of conversion of MCI to AD. However, the unprecedented scale and complexity of these neuroimaging genomic data sets have presented critical computational challenges for achieving the full transformative potential from comprehensive joint analysis of these heterogeneous and longitudinal data sets. This project aims to address these emerging challenges for early prediction of MCI-to-AD conversion with four aims. Aim 1 is to develop a novel sparse bi-multivariate learning model based system biology framework for analysis of genome-wide association results across a large number of the structural and functional phenotypes derived from neuroimaging scans of the whole brain. Our new methods are designed for bi-multivariate analysis of high-throughput genomic data and complex Quantitative Traits (QTs) related to MCI-to-AD conversion by utilizing the system biology knowledge. Aim 2 is to further extend the learning approaches in Aim 1 with the new structured sparse models to the multi-dimensional data integration methods to identify the heterogeneous biomarkers from multiscale interrelated imaging genomic data for outcome prediction. Meanwhile, we will utilize the joint multi-task learning scheme to identify the stable genetic and phenotypic biomarkers that are associated with cognitive functions decline and MCI-to-AD conversion simultaneously. Based on the studies in Aims 1 and 2, Aim 3 is focused on revealing the longitudinal biomarkers of the changes of MCI progression or cognitive impairment by a new structured low-rank multi-task regression model. These biomarkers can fully differentiate longitudinal profiles of relevant QTs and better capture genetic associations with QT changes over time. Aim 4 is to evaluate and validate our proposed machine learning and bioinformatics algorithms on both synthetic data and real imaging genomic data. The results of this project will be able to efficiently improve our understanding of the complex neurobiological mechanism underlying the MCI-to-AD conversion. The identified biomarkers will finally enhance the early and accurate prediction of MCI-to-AD conversion such that the clinical treatment can be provided in time.

The research objective of this proposal is to address the computational challenges in an innovative BIGDATA application on imaging-omics based precision medicine. Recent advances in high-throughput imaging (such as histopathology image) and multi-omics (such as DNA sequence, RNA expression, methylation, etc.) technologies created new opportunities for exploring relationships between histology, molecular events, and clinical outcomes using quantitative methods. However, the unprecedented scale and complexity of these imaging-omic data have presented critical computational bottlenecks requiring new concepts and enabling tools. This project builds a new computational framework to integrate novel big data mining algorithms with cloud and high-performance computing strategies for revealing complex relationships between histopathology images, multi-omics, and phenotypic outcomes. This project is innovative and crucial not only to facilitating the development of new big data mining techniques, but also to addressing emerging scientific questions in imaging-omics and many other biomedical applications. The developed methods and tools are expected to impact other cancer genomics research and enable investigators working on cancer medicine to effectively test their scientific hypothesis. This project facilitates the development of novel educational tools to enhance several current courses. University of Texas at Arlington is a minority-serving institution and has large population of Hispanic and Black Americans. This project engages the minority students and under-served populations in research activities to give them a better exposure to cutting-edge science research.

To solve the key and challenge problems in big imaging-omics data mining, this project explores the following research tasks. First, the large-scale non-convex sparse learning models are developed for identifying outcome-relevant phenotypic traits from big histopathology images. Second, the biological domain knowledge is utilized to guide the sparse learning models to uncover the molecular bases of complex traits. Third, the data integration models are designed to integrate imaging-omics data from multiple sources and discover the heterogeneous biomarkers. Fourth, the Baysian learning model is explored to predict longitudinal cancer outcomes. Fifth, the cloud computing and high-performance computing strategies are developed to support the big imaging-omics data mining, such as optimizations for various data mining workloads on heterogeneous hardware (e.g. GPU and NUMA multicore processors) to fully unlock the potential of data center hardware. It is innovative to integrate big data mining algorithms with cloud and high-performance computing to imaging-omics that hold great promise for a systems biology of the precision medicine.

This project investigates new robust large-scale data mining and machine learning algorithms to solve critical computational challenges in mining massive depression thought records for cognitive behavior therapy. Depression is rapidly emerging as one of the major problems in our society and is also related to many other health conditions, such as stroke, diabetes, hypertension, HIV/AIDS, etc. Cognitive behavior therapy is the most extensively researched form of psychotherapy for depression, and the depression thought records from patients is the key component of cognitive behavior therapy. However, the process of reviewing and analyzing the depression thought records is extremely time consuming, which inhibits both clinical interviews and the training of new therapists. This project builds a novel data mining system to automatically discover knowledge from depression thought records for assisting therapists in selecting potential interventions and aiding new therapists in their development of cognitive behavior therapy skills. This project will facilitate the development of novel educational tools to enable new courses and enhance current courses. This project engages minority students and under-served populations in research activities to give them a better exposure to cutting-edge science research.

To effectively and efficiently analyze large-scale depression thought records, this project explores the following research tasks. First, the project develops a robust semi-supervised learning model to categorize logical thinking errors of depression thought records. Second, the project investigates a joint multi-task method to simultaneously recognize the categories of thinking errors and emotions of depression thought records. Third, new multi-label and multi-instance learning is studied for identifying coping activities. Fourth, to analyze the multi-language depression thought records, robust transfer learning methods are developed for cross-language knowledge transfer. Meanwhile, parallel computational algorithms are designed and applied for large-scale depression thought record data mining. These novel data mining algorithms are designed to solve large-scale applications and automate the depression thought record data mining, which holds great promise for smart health.

With advances in anesthesia techniques, surgery has become increasingly applicable to a wider range of diseases and patients. Worldwide more than 230 million major surgical procedures are carried out each year. In terms of patient safety and medical economics, an important issue is how to reduce the incidence of postoperative complications and mortality. At least half of postoperative complications can be prevented, while improvements in anesthesia-associated factors contribute greatly to the prevention of complications. Anesthesia information management system is a specialized type of electronic health record that allow the automatic and reliable collection and storage of patient data during the perioperative period. The electronic anesthesia data not only provide a rich data set to assist both anesthesia providers and hospitals with their goals to improve patient safety during the fast-paced intra-operative period, but also capture detailed data to allow end users to access information for management, quality assurance, and research purposes. This project addresses the computational challenges in large-scale electronic anesthesia data mining, develops and validates an automated anesthesia risk prediction and decision support system to identify risk factors and detect patients at risk of postoperative complications and in-hospital mortality.

This project develops novel large-scale machine learning framework to integrate the emerging key computational techniques, such as semi-supervised generative adversarial learning, interpretable deep learning, large-scale optimization, and unsupervised hashing, to analyze large-scale electronic anesthesia data for enhancing anesthesia risk stratification and improving the quality of care for precision health. Specifically, the PIs investigate: 1) new computational tools to automate electronic anesthesia data processing, 2) novel semi-supervised generative adversarial network for anesthesia risk stratification, 3) interpretable deep learning model for clinical markers discovery, 4) scale up deep learning models for big data computation via new large-scale optimization algorithms, 5) new unsupervised deep generative adversarial hashing network for fast and accurate clinical case retrieval, and 6) evaluate the proposed methods and system using real large-scale anesthesia data. It is innovative to integrate large-scale machine learning and data-intensive computing for electronic anesthesia data mining that holds great promise for predicting postoperative outcomes using the comprehensive preoperative and intra-operative patient profiles. The developed methods and tools impact other public health research and enable investigators working on electronic health data to effectively test risk prediction hypothesis. This project facilitates the development of novel educational tools to enhance several current courses at University of Pittsburgh.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Recent advances in multimodal brain imaging and high throughput genotyping and sequencing techniques provide exciting new opportunities to ultimately improve our understanding of brain structure and neural dynamics, their genetic architecture, and their influences on cognition and behavior. However, data privacy and security issues have inhibited data sharing across institutes. Emerging multi-site collaborative data analysis can address these issues and facilitate data and computing resource sharing. In collaborative data analysis, the participating institutes keep their own data, which are analyzed and computed locally, and only share the computed results by communicating with a server. The server communicates with all institutes and updates the local models such that the trained machine learning models indirectly use all data and are shared with all institutes. Although some distributed/parallel computation techniques were recently proposed to address big data mining problems, most of them are synchronous models. Asynchronous distributed learning methods are much more efficient, because they allow the server to update the model with information from only one worker node without waiting for slow worker nodes in each round. However, the convergence analysis for the asynchronous distributed algorithms is much more difficult due to the inconsistent variables update across nodes. Thus, it is challenging to design efficient distributed machine learning algorithms for collaborative big data analysis. The research objective of this project is to address the computational challenges in the emerging multi-site collaborative data mining for brain big data.

This project seeks to harness the opportunities of designing new efficient asynchronous distributed machine learning algorithms with rigorous theoretical foundations for multi-site collaborative brain big data mining, creating large-scale computational strategies and effective software tools to reveal sophisticated relationships among heterogeneous brain data. This project designs the asynchronous distributed machine learning and principled big data mining models to conduct the comprehensive study of brain imaging genomics and connectomics. Specifically, the principal investigators investigate: 1) collaborative genotype and phenotype association study using new asynchronous doubly stochastic proximal gradient algorithms; 2) communication-efficient multi-site collaborative data integration models to integrate imaging genomics data for predicting outcomes of interest; 3) collaborative deep learning algorithm speedup by the asynchronous distributed algorithms with applications in temporal cognitive change prediction; and 4) new graph convolutional deep learning models for brain network mining. It is innovative to integrate new distributed machine learning and data-intensive computing with brain imaging genomics and connectomics that hold great promise for a systems biology of the brain. The developed methods and tools impact other neuroimaging, genomics, and neuroscience research, and enable investigators working on brain science to effectively test their scientific hypotheses. This project will also facilitate the development of novel educational tools.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

A key to combat the Coronavirus Disease (COVID-19) pandemic is to prevent the pandemic from overloading the public healthcare system, so that sufficient medical resources could be available for hospitalized patients. This project will develop new mobile sensing and Artificial Intelligence (AI) techniques for in-home evaluation of COVID-19 infection in order to pursue automated and non-invasive screening of potential viral disease carriers. It aims to timely identify negative cases caused by other diseases with similar symptoms, and hence avoids unnecessary hospital visits as many as possible.

The proposed techniques will use commodity smartphones to measure the changes of humans? airway mechanics, which are uniquely correlated to COVID-19 infection. These measurements build on acoustic sensing with smartphones? built-in speakers and microphones. First, new acoustic waveforms will be designed to minimize acoustic signal distortion in human airways. Second, new signal processing techniques will be developed for accurate measurements. Third, deep learning techniques will be used to develop generic models that depict the core characteristics of airway mechanics. These techniques will be evaluated by lab testing and experiments with student volunteers. This research will enable identifying false positives of COVID-19 infection out of the clinic and could contribute to the containment of the virus spread and damage. The proposed technologies will be applicable to a wide variety of commodity smartphones and could also be used in daily practice with handmade mouthpieces. Broader impacts will also result from a variety of education and outreach activities. New courses will be developed to incorporate the outcome of this research, and the research outcome will be disseminated through technology transfer to industry. The outcome of this project, including source codes and collected data from student volunteers, will be maintained at the project repository (http://www.pitt.edu/~weigao/research_COVID19.html) for at least five years, and will be made available to the public community.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

ABSTRACT In response to PAR-19-269 ?Cognitive Systems Analysis of Alzheimer's Disease Genetic and Phenotypic Data (U01 Clinical Trial Not Allowed)?, our project unites experts in AD genomics, machine learning and AI (including deep learning), large-scale data integration, and international data harmonization to work in a carefully-designed Consortium Structure in close partnership with the NIH, ADSP, and NIAGADS. We will develop a suite of complementary big data analytic approaches for ultra-scale analysis of Alzheimer?s Disease (AD) genomic and phenotypic data. The vast data volumes now generated by the Alzheimer?s Disease Sequencing Project (ADSP), National Alzheimer?s Coordinating Center (NACC), Alzheimer?s Disease Neuroimaging Initiative (ADNI), Accelerating Medications Partnership AD (AMP-AD), and UK Biobank (UKBB), far exceed the capacity of all current analytic methods, which have not kept pace with the scale and speed of data collection. This vast amount of genetic and phenotypic data mandates new and more powerful algorithms to: (1) store, manage, and manipulate whole-genome sequences and associated data on an ever-growing scale; (2) discover novel AD risk and protective loci by merging informatics and AD genomics databases; (3) relate whole-genome changes to the ATN(v) biomarkers that now define biological AD. Our Ultrascale Machine Learning Initiative, or ?ULTRA? - will offer new AI and deep learning tools to discover features in massive scale genomics data - relating whole genome data to biomarker features by merging all relevant data sources. Our team of experienced PIs will coordinate efforts across the U.S. to create these large-scale data analytic tools. Our MPI team and 6 Core Leads have decades of experience working together and with the AD community in pioneering machine learning methods for AD genetics and neuroimaging, including leadership of international neuroimaging consortia across the world. Dedicated Cores focus on Genomic, Imaging, and Cognitive Data Harmonization. Curated data will then be efficiently imported into AI approaches and informatics pipelines that will allow the AD research community to leverage ultra-scale, multidimensional genomic and phenotypic data from the ADSP, NACC, ADNI, AMP-AD, and others. Our work is organized by a carefully-designed and coordinated Consortium guided by all stake-holders, clinical leaders, and pioneering analysts in AD genomics and neuroimaging. Our ultrascale AI tools will advance AD genomics research and will include efforts in training, and a dedicated Drug Repurposing Core. This team effort will accelerate understanding of the genetic, molecular and neurobiological mechanisms of AD, yielding significant translational impact on disease and drug development.

Advances in machine learning have made a major impact on many real-world applications over the past decade, and have achieved scientific and engineering breakthroughs across many disciplines. A new era of collaborative learning is emerging as part of the next phase of ubiquitous computing, wherein researchers at different sites will work together to correlate the disparate data they have separately acquired and eventually create a sophisticated decision-making model. It is thus imperative to establish a platform to support collaborative, multi-party data analysis, through which the participating parties can share their data with each other with different degrees of privacy control. The participants can compute with each other's data, by either directly sharing data with the server or only sharing their model parameters with the server to collaboratively derive a solution with other parties. To make such an environment available to the community, this project establishes a scalable and trusted hardware and software environment, termed Bridge, to support a general form of collaborative machine learning. The Bridge platform enables scalable multi-party learning and data analysis in a variety of forms, in both centralized and decentralized settings, with security and privacy guarantees. The project's novelties are to synergistically design and integrate both hardware and software innovation as well as a suite of security and privacy mechanisms and tools to support various types of multi-party machine learning. The project's impacts are to enable collaborative research efforts in diverse communities of CISE researchers pursuing focused research agendas in computer and information science and engineering, and generate enormous social and economic benefits to individuals and organizations. The minority students and under-served populations will be engaged in research activities to create an inclusive environment where everyone contributes to and benefits from cutting-edge scientific research.<br/><br/><br/>The Bridge platform will develop a unified hardware and software infrastructure to achieve hardware and software co-design for multi-party learning. An algorithmic software infrastructure is designed to support distributed, federated, and multi-modal model learning and sharing. The Bridge platform integrates cryptographic (secure multi-party computation) and noise-based methods (differential privacy) to provide privacy across the entire process from data collection to output. The Bridge platform provides a set of tools on integrated data access, AutoML, team creation, machine learning model vulnerability evaluation, and heterogeneous feature embeddings to support flexible user applications. The Bridge platform ensures the scalability in the number of tasks, the number of users, and heterogeneity of data types by developing advanced techniques to improve asynchronous model updates, communication efficiency, fast convergence, and vertical data partition. The Bridge platform builds a collaborative learning community and accelerates many new research areas in the core Computer and Information Science and Engineering (CISE), such as advanced machine learning and data science, data privacy and trustworthy AI, convergent research among hardware, software and machine learning, and intelligent internet of things.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

The research objective of this proposal is to address the computational challenges in the innovative nanomaterial data analysis or nanoinformatics for predicting nanomaterials properties. Nanomaterials are very small materials that can be used in a variety of applications, including nanomedicine development. The vast quantities of existing experimental data require new nanoinformatics approaches and toolkits for data extraction, analysis, and sharing. This can help guide the safe design of next-generation of nanomedicines with desirable therapeutic activities, while also ensuring they have limited side effects. However, there are currently two critical limitations to using machine learning approaches in nanoinformatics modeling studies. First, most existing data available for modeling were based on a limited number of nanomaterials that also have limited experimental characterization of their chemical properties. Second, despite significant efforts from various researchers, the available modeling approaches that have been developed are applicable only for a specified small set of nanomaterials and have rarely been used to design nanomaterials. This project will address the computational challenges in large-scale nanomaterial data mining, development and validation of an automated informatics framework to digitalize nanostructures, identify molecular markers, and support fast nanomaterial retrieval and integrative analysis. This project will also facilitate the development of novel educational tools to enhance several current courses at Rutgers University, University of Pittsburgh, and University of Minnesota. The investigators will engage the minority students and under-served populations in research activities to give them a better exposure to cutting-edge science research.<br/><br/>In this project, a novel machine learning based nanoinformatics framework will be developed to integrate new digital nanostructure representations with the emerging key computational techniques. The project focuses on designing principled machine learning and data science algorithms for analyzing large-scale nanomaterial data to create new informatics toolkits to facilitate the nanomedicine-based treatments and new nanomaterial design. Specifically, the following research goals will be met in this project: 1) new computational tools to automate nanostructure digitalization; 2) interpretation method to enhance deep learning based predictive models; 3) new cross-modal deep hashing network for fast and accurate nanomaterial data retrieval; and 4) evaluate the proposed methods and system using real large-scale nanomaterial data and release the database and nanoinformatics toolkits to the public. Unlike most existing nanoinformatics strategies that perform modeling and analysis at a small scale, this project will provide promising new directions to the analysis of large-scale complex nanomaterial data by addressing the critical data-intensive analysis issues including efficiency, scalability, and interpretability. The investigations combine rigorous theoretical analysis and emerging application studies and will contribute to both academic research and potential commercialized products. This project will advance and thus extend the relationship between engineering innovation and computational analysis, and hold great promise for nanomaterial and nanomedicine developments.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Pulmonary diseases constitute a major public health challenge, and pulmonary function testing (PFT) is the main method of evaluating the changes in human airway mechanics, which are the key symptoms of pulmonary diseases. Due to the possibility of frequent exacerbations in pulmonary diseases, it is important that PFT is accessible to patients anytime and anywhere out of clinic, but most of current in-clinic PFT techniques are too cumbersome and expensive to be used out of clinic. This project addresses these challenges to enable PFT anytime and anywhere, by developing new and integrated artificial intelligence (AI) and sensing systems on commodity smartphones. As pulmonary diseases widely affect the human society and result in billions of annual healthcare related costs, this project has great potential to benefit society by enabling efficient disease monitoring, diagnosis and management out of clinic. This project is also contributing to society by producing datasets that help understand the disease mechanisms, as well as developing new curricula, disseminating research for education and training, and engaging underrepresented students in research.<br/><br/>The primary goal of this project is to enable highly accurate, adaptable and generic PFT out of clinic using commodity smartphones. The project consists of four research tasks: (1) designing new acoustic sensing systems on commodity smartphones that measure the humans’ airway lumen dimensions and characteristics; (2) extracting appropriate biomarker profiles from acoustic sensory data for disease evaluation; (3) developing generalizable machine learning (ML) models that can be applied to evaluating different pulmonary diseases; (4) exploring distributed and asynchronous methods of training the ML models with new federated learning techniques.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

The newly emerging Artificial Intelligence (AI) of Things (AIoT) and Internet of Senses (IoS) systems will make mobile and embedded devices smart, communicative, and powerful by processing data and making intelligent decisions through the integration of the Internet of Things (IoT) and Artificial Intelligence (AI). This project aims to provide a new generation of systems, algorithms, and tools to facilitate such deep integration at extreme scale. The novelty of the project is to fundamentally ensure scalability of future Machine Learning (ML) systems over the large population of distributed devices, by formulating the seamless integration of advanced ML algorithms with co-designed hardware, computer architectures, and distributed edge-cloud systems, along with meaningful security and privacy guarantees. This co-design methodology allows synergistic consideration of the intrinsic heterogeneity, performance and energy constraints of devices, as well as the unprecedented scale and complexity of data produced by these devices. The project's impacts are to lay the foundation for the future of AIoT and IoS systems by solving challenges driven by needs related to their complex and heterogeneous contexts, and to advance a wide swath of fields including ML, edge computing, IoT, hardware, software and related engineering disciplines. This project is also contributing to society through developing new curricula, disseminating research for education and training, engaging under-represented students in research, and outreaching to high-school students.<br/><br/>The primary goal of this project is to build a new co-designed framework of hardware, software, and algorithms to enable extreme-scale ML systems for the emerging AIoT and IoS systems. The project consists of five research thrusts. Thrust 1 develops hardware, computer architecture and compiler approaches to address the scalability issue in AIoT and IoS systems by enforcing large-scale split learning on devices. Thrust 2 investigates extreme-scale ML on weak embedded devices by designing a new system framework that adaptively partitions and offloads the ML computing workloads. Thrust 3 addresses system and data unreliability by designing new cross-layer algorithms and hardware techniques. Thrust 4 investigates algorithm, hardware and software co-design to enable secure and privacy-preserving ML systems at scale. Thrust 5 involves designing and implementing an IoS testbed and a smart building testbed to evaluate the proposed system designs.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Recent developments of single-cell omics technologies enable multi-modality measurements at genome, transcriptome, epigenome, or proteome scale, which will lead to unprecedented insight and resolution to fundamental biological processes. The project will construct a novel bioinformatics framework with advanced machine learning models, efficient computational tools, and user-friendly software for single-cell multi-omics data analysis. The outputs will be available online to the public and are expected to impact biological research community and empower scientists working on single-cell data to effectively test biological hypothesis, especially knowledge extraction from massive high-dimensional and complex datasets. The project will facilitate the development of novel educational tools to enhance curriculum design. Minority students and under-served populations will be engaged in cutting-edge research activities.<br/> <br/>The project focuses on designing principled machine learning and bioinformatics algorithms for analyzing large-scale single-cell multi-omics data to create toolkits to facilitate biological research. Specially, the research team will investigate 1) new cross-modal deep canonical correlation self-supervised autoencoder for multi-modal single-cell data integration, 2) new computational methods to study the associations of single-cell RNA-seq data and protein markers via semi-supervised deep neural networks, 3) interpretation algorithms to enhance predictive model via utilizing structure semantic information and identified biomarkers, 4) statistical inference framework for identifying and inferring conditional dependence from single-cell data, 5) novel transformer based variational autoencoder model for super-resolution spatial transcriptomics, 6) tool portal development for single-cell data analysis to advance biology research, and 7) validations of the proposed methods and system using real large-scale single-cell data. The project is innovative in integrating large-scale machine learning and data-intensive computing for single-cell bioinformatics and will hold great promise for biological mechanism understanding and biomedicine development. The results of the project can be found at: https://sites.pitt.edu/~heh45/NSF2225775.html<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.