Timezone: »
Complex nonlinear time-series data can be effectively modeled by Switching Linear Dynamical System (SLDS) models. In trying to allow for unbounded complexity in the discrete modes, most approaches have focused on Dirichlet Process mixture models. Such non-parametric Bayesian models restrict the distribution over dynamical modes to be exchangeable, making it difficult to capture important temporally and spatially sequential dependencies. In this work, we address these concerns by developing the non-exchangeable SLDS (neSLD) model class effectively extending infinite-capacity SLDS models to capture non-exchangeable distributions over dynamical mode partitions. Importantly, from this non-exchangeability, we can learn transition probabilities with infinite capacity that depend on observations or on the continuous latent states. We leverage partial differential equations (PDE) in the modeling of latent sufficient statistics to provide a Markovian formulation and support efficient dynamical mode updates. Finally, we demonstrate the flexibility and expressivity of our model class on synthetic data.
Author Information
Victor Geadah (Princeton University)
Jonathan Pillow (Princeton University)
More from the Same Authors
-
2021 : Neural Latents Benchmark ‘21: Evaluating latent variable models of neural population activity »
Felix Pei · Joel Ye · David Zoltowski · Anqi Wu · Raeed Chowdhury · Hansem Sohn · Joseph O'Doherty · Krishna V Shenoy · Matthew Kaufman · Mark Churchland · Mehrdad Jazayeri · Lee Miller · Jonathan Pillow · Il Memming Park · Eva Dyer · Chethan Pandarinath -
2022 Poster: Dynamic Inverse Reinforcement Learning for Characterizing Animal Behavior »
Zoe Ashwood · Aditi Jha · Jonathan Pillow -
2022 Poster: Extracting computational mechanisms from neural data using low-rank RNNs »
Adrian Valente · Jonathan Pillow · Srdjan Ostojic -
2020 Poster: High-contrast “gaudy” images improve the training of deep neural network models of visual cortex »
Benjamin Cowley · Jonathan Pillow -
2020 Poster: Identifying signal and noise structure in neural population activity with Gaussian process factor models »
Stephen Keeley · Mikio Aoi · Yiyi Yu · Spencer Smith · Jonathan Pillow -
2020 Poster: Inferring learning rules from animal decision-making »
Zoe Ashwood · Nicholas Roy · Ji Hyun Bak · Jonathan Pillow -
2018 Poster: Scaling the Poisson GLM to massive neural datasets through polynomial approximations »
David Zoltowski · Jonathan Pillow -
2018 Poster: Efficient inference for time-varying behavior during learning »
Nicholas Roy · Ji Hyun Bak · Athena Akrami · Carlos Brody · Jonathan Pillow -
2018 Poster: Model-based targeted dimensionality reduction for neuronal population data »
Mikio Aoi · Jonathan Pillow -
2018 Poster: Power-law efficient neural codes provide general link between perceptual bias and discriminability »
Michael J Morais · Jonathan Pillow -
2018 Poster: Learning a latent manifold of odor representations from neural responses in piriform cortex »
Anqi Wu · Stan Pashkovski · Sandeep Datta · Jonathan Pillow -
2017 Poster: Gaussian process based nonlinear latent structure discovery in multivariate spike train data »
Anqi Wu · Nicholas Roy · Stephen Keeley · Jonathan Pillow -
2016 Poster: Bayesian latent structure discovery from multi-neuron recordings »
Scott Linderman · Ryan Adams · Jonathan Pillow -
2016 Poster: Adaptive optimal training of animal behavior »
Ji Hyun Bak · Jung Choi · Ilana Witten · Athena Akrami · Jonathan Pillow -
2016 Poster: A Bayesian method for reducing bias in neural representational similarity analysis »
Mingbo Cai · Nicolas W Schuck · Jonathan Pillow · Yael Niv -
2015 Poster: Convolutional spike-triggered covariance analysis for neural subunit models »
Anqi Wu · Il Memming Park · Jonathan Pillow