Bulletin of the American Physical Society
APS March Meeting 2016
Volume 61, Number 2
Monday–Friday, March 14–18, 2016; Baltimore, Maryland
Session S55: Inference in BiophysicsInvited
|
Hide Abstracts |
Sponsoring Units: DBIO GSNP Chair: Steve Presse, Indiana University of Pennsylvania Room: Hilton Baltimore Holiday Ballroom 6 |
Thursday, March 17, 2016 11:15AM - 11:51AM |
S55.00001: Genetic networks and the flow of positional information in embryonic development Invited Speaker: William Bialek When we study a biological system, we make inferences about the underlying mechanisms and dynamics. But biological systems themselves must also solve inference problems, as when our brains draw conclusions about the world given (often quite limited) data from our eyes and ears. My colleagues and I have been exploring both of these inference problems as they play out in the first hours of development in the fruit fly embryo. In this system, the concentrations of particular molecules encode the position of each cell in the embryo, and these concentrations are the outputs of a genetic network. Putting ourselves in the place of the cells, we have been able to read the code, building a dictionary that maps gene expression levels back into estimates of position. If our dictionary really is the one used by the embryo, then mutants should build predictably distorted body plans, and preliminary results show quantitative agreement with these predictions. Independent of their role as carriers of information, we can also analyze the patterns of gene expression to draw inferences about the underlying network. Finally, it is possible that the network architecture and parameters have been chosen to optimize the flow of information, and we see signatures of this optimization. Joint work with CG Callan, JO Dubuis, T Gregor, D Krotov, M Petkova, TR Sokolowski, G Tkacik, AM Walczak, and EF Wieschaus. [Preview Abstract] |
Thursday, March 17, 2016 11:51AM - 12:27PM |
S55.00002: Deep Learning, Group representations, and the Information-Bottleneck phase transitions. Invited Speaker: Naftali Tishby Deep Neural Networks (DNNs) are analyzed via the theoretical framework of the information bottleneck (IB). We first show that any DNN can be quantified by the mutual information between the layers and the input and output variables. Using this representation we can calculate the optimal information theoretic limits of the DNN and obtain finite sample generalization bounds. The advantage of getting closer to the theoretical limit is quantifiable both by the generalization bound and by the network's simplicity. We argue that both the optimal architecture, number of layers and features/connections at each layer, are related to critical points on the information bottleneck tradeoff line, namely, relevant compression of the input layer with respect to the output layer. The hierarchical representations at the layered network naturally correspond to the structural phase transitions along the information curve. An interesting class of solvable DNN's arise by applying this framework to the case of~symmetries in the~supervised learning task. The case of~translation invariance leads to the familiar convolution neural networks. Other symmetry groups yield~different types of bifurcation diagrams and network architectures, which correspond to information contained by irreducible representations of the group. These new insights also suggest new sample complexity bounds, architecture design principles (number and widths of layers), and eventually entirely different deep learning algorithms.\\ \\Based partly on works with Noga Zaslavsky and Ravid Ziv. [Preview Abstract] |
Thursday, March 17, 2016 12:27PM - 1:03PM |
S55.00003: Fock spaces for modeling macromolecular complexes Invited Speaker: Justin Kinney Large macromolecular complexes play a fundamental role in how cells function. Here I describe a Fock space formalism for mathematically modeling these complexes. Specifically, this formalism allows ensembles of complexes to be defined in terms of elementary molecular ``building blocks’’ and ``assembly rules.’’ Such definitions avoid the massive redundancy inherent in standard representations, in which all possible complexes are manually enumerated. Methods for systematically computing ensembles of complexes from a list of components and interaction rules are described. I also show how this formalism readily accommodates coarse-graining. Finally, I introduce diagrammatic techniques that greatly facilitate the application of this formalism to both equilibrium and non-equilibrium biochemical systems. [Preview Abstract] |
Thursday, March 17, 2016 1:03PM - 1:39PM |
S55.00004: Causal inference of signaling networks using single cell data Invited Speaker: Karen Sachs |
Thursday, March 17, 2016 1:39PM - 2:15PM |
S55.00005: Maximum Entropy and the Inference of Pattern and Dynamics in Ecology Invited Speaker: John Harte Constrained maximization of information entropy yields least biased probability distributions. From physics to economics, from forensics to medicine, this powerful inference method has enriched science. Here I apply this method to ecology, using constraints derived from ratios of ecological state variables, and infer functional forms for the ecological metrics describing patterns in the abundance, distribution, and energetics of species. I show that a static version of the theory describes remarkably well observed patterns in quasi-steady-state ecosystems across a wide range of habitats, spatial scales, and taxonomic groups. A systematic pattern of failure is observed, however, for ecosystems either losing species following disturbance or diversifying in evolutionary time; I show that this problem may be remedied with a stochastic-dynamic extension of the theory. [Preview Abstract] |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700