2017 — 2022 |
Palmer, Stephanie |
N/AActivity Code Description: No activity code was retrieved: click on the grant title for more information |
Career: Understanding Vision and Natural Motion Statistics Through the Lens of Prediction
The visual input to the brain is transformed even before signals leave the eye, and these computations produce an efficient representation of the structure of the natural visual world. Previous work by the PI has shown that this processing can include repackaging of information for optimal prediction. This suggests a new approach to neural encoding. While many previous studies have sought to characterize what stimuli in the past gave rise to a subsequent response, this work asks what future stimuli those responses predict. The proposed project will derive the best possible predictor given the way objects move in the outside world and quantify how close the brain gets to this optimum. Viewing the brain through the lens of prediction develops a principle of neural coding and computation that can bridge brain regions, from the retina to higher visual areas. A component of this plan involves measuring and quantifying the predictive components of natural motion. In doing so, a public database of natural motion will be created that will be a lasting tool for the neuroscience and computer vision communities. An associated educational program will bring over 100 local middle school children to campus each year for hands-on neuroscience experiments, and will instill in a large group of graduate students the rewards and responsibilities of science teaching.
The research proposed here explores prediction in the visual system in a variety of ways: by computing efficiency bounds on the predictive encoding of complex motion, by developing quantitative methods to test these bounds in neural datasets, by measuring the statistics of motion in natural scenes, and by describing how, mechanistically, the brain achieves this performance. Hypotheses about how the brain performs optimal predictive computations may be constrained by the structure of predictable events in the natural visual world. To measure these statistics, a new natural movie database will be constructed by making high-speed, high-pixel-depth recordings of natural scenes. By quantifying motion in these data, this project will yield statistical and generative models of natural motion that will inform our understanding of the natural world and provide a compact way to recapitulate natural motion in silico. These stimuli will be used to test whether neural systems optimally encode information relevant for prediction. The work will also test what adaptive and otherwise non-linear processing steps underlie optimal prediction in the brain.
|
0.915 |
2017 — 2020 |
Allesina, Stefano (co-PI) [⬀] Prince, Victoria [⬀] Palmer, Stephanie |
N/AActivity Code Description: No activity code was retrieved: click on the grant title for more information |
Nrt-Ige: Reproducibility and Rigor in Quantitative Biology: a Hands-On Approach
Current research in biology is producing increasingly large and complex sets of data. These data could represent, for example: DNA sequences, images of the brain, or number of species in an ecosystem. In each case, unlocking the information within these big data sets requires sophisticated mathematical and computational approaches. The standard curriculum for graduate students in biological sciences was designed well before this data deluge. As a consequence, today's graduate students are not being adequately trained for their future careers. At the same time, there are growing concerns that scientists are sometimes unable to reproduce published findings. This inability often results from poor data analysis strategies. The future success of the US biological research mission hinges on training students to use data analysis approaches that are both rigorous and reproducible. This National Science Foundation Research Traineeship (NRT) award in the Innovations in Graduate Education (IGE) Track to the University of Chicago seeks to meet this need by developing a new and effective approach to the training of early stage graduate students in the quantitative analysis of biological data.
The overarching goal of this program is to teach students to critically evaluate quantitative analysis methods in the scientific literature, and to acquire good programming habits that support reproducibility and rigor in their own research. An interdisciplinary team of quantitative biologists will direct and lead the program, exposing students to the faculty that can advise them in future work. The training program begins with an intensive residential week-long boot camp that brings together students across diverse sub-fields of biology to promote teamwork and prepare them for interdisciplinary research. The boot camp includes introductory tutorials in computer programming, statistics, and modeling in modern biology, as well as more advanced tutorials in statistical approaches to large data sets and practical lessons in organizing and sharing code and data. The boot camp is capped off with a series of workshops in which students apply what they have learned to real biological data spanning a wide range of fields. A subsequent on-campus course builds on and reviews these concepts, and integrates training in rigor and reproducibility with concepts of responsible research. We hypothesize that this program will produce trainees who are well-prepared for the future scientific workforce. We will evaluate the impact of this intervention through quizzes, surveys, and targeted interviews. All teaching materials and data sets used in the workshops will be shared online so that any university can implement a similar training module on their own campus.
The NSF Research Traineeship (NRT) Program is designed to encourage the development and implementation of bold, new, potentially transformative models for STEM graduate education training. The Innovations in Graduate Education Track is dedicated solely to piloting, testing, and evaluating novel, innovative, and potentially transformative approaches to graduate education.
|
0.915 |
2020 |
Bialek, William (co-PI) [⬀] Palmer, Stephanie E Schwab, David Jason (co-PI) [⬀] |
R01Activity Code Description: To support a discrete, specified, circumscribed project to be performed by the named investigator(s) in an area representing his or her specific interest and competencies. |
Coarse-Graining Approaches to Networks, Learning, and Behavior
Project Summary The theory hub put forward in this proposal will work to translate successful and powerful approaches to describing emergent collective behavior in physical systems so they can be applied to the brain. Working closely together, the three theorists will develop methods for finding and quantifying the relevant modes of population activity in the brain, both in instantaneous snapshots of activity and activity as it evolves in time. Methods will be tested in a wide range of neural systems at different processing stages and scales: from salamanders to rodents to humans, from the retina to the cortex, from tens to thousands of cells. The approach will be validated by checking that the neural code can be read out with high fidelity even after being compressed into a much smaller subspace. The project will produce data analysis code that will be made available for neuroscience researchers to use on their own data, in addition to the results of the analyses of the particular systems studied. The neural code is inherently collective; while single neurons execute sophisticated computations, hundreds to thousands of neurons are utilized to sense the environment and drive behavior in even the simplest organisms. Although the past hundred years have yielded substantial progress in neuroscience, only recently have researchers had the capacity to record from complete neural populations - that is, to view the collective behavior of a functioning neural network. With these rapid experimental advances, there is an urgent need for complementary theoretical and computational approaches to guide the exploration of emergent behavior in large groups of neurons, allowing one to turn `big data' into `big ideas'. This proposal outlines a path towards a new theoretical framework for finding and quantitatively analyzing collective phenomena in the brain that underlie sensory coding, the representation of space, prediction, and ultimately drive behavior. The project draws heavily on the success of so-called renormalization group approaches in theoretical physics that revolutionized the understanding of collective phenomena in physical systems, and sculpted much of the progress in statistical physics in the second half of the twentieth century. The methods explored in this proposal generalize such techniques so they can be applied to a much wider range of problems. The methods developed by this theory hub based on the renormalization group will be applicable to a wide range of neural data since they are explicitly designed to generalize techniques from theoretical physics to a much broader setting. Indeed, a larger goal of the approach is to search for universality in collective behavior in the neural code. The techniques proposed are relatively straightforward to execute and will provide a fundamental methodology for interrogating high-dimensional data in fields as diverse as behavioral neuroscience and biophysics. The new techniques will also be taught as part of the three theorists' ongoing efforts to expose incoming graduate students in biological sciences to quantitative methods in biology.
|
1 |