Information Geometry of the Retinal Representation Manifold

Authors: Xuehao Ding, Dongsoo Lee, Joshua Melander, George Sivulka, Surya Ganguli, Stephen Baccus

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To model the joint probability distribution of neural responses conditioned on the stimulus, we created a stochastic encoding model of a population of salamander retinal ganglion cells based on a three-layer convolutional neural network model. This model not only accurately captured the mean response to natural scenes but also a variety of second-order statistics. With the model and the proposed theory, we computed the Fisher information metric over stimuli to study the most discriminable stimulus directions.
Researcher Affiliation Academia 1,5Department of Applied Physics, Stanford University 2,3Neurosciences Phd Program, Stanford University 4Department of Electrical Engineering, Stanford University 6Department of Neurobiology, Stanford University {xhding, dsnl, melander, gsivulka, sganguli, baccus}@stanford.edu
Pseudocode No The paper includes a diagram of the model architecture (Figure 2) and describes the methods in paragraph form, but it does not present any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any links to source code repositories or explicitly state that the code for the described methodology is publicly available.
Open Datasets Yes The deterministic part of the model including its hyperparameters and training method is adopted from Ref. [12, 13, 14]
Dataset Splits No The paper mentions data optimization and evaluation, including a 'test set' for results, but it does not specify explicit training, validation, and test dataset splits with percentages, sample counts, or specific predefined split methodologies.
Hardware Specification Yes The optimization was performed using ADAM [25] via Pytorch [26] on NVIDIA TITAN Xp, Ge Force GTX TITAN X, Ge Force RTX 3090, TITAN RTX, TITAN V, and Ge Force RTX 2080 Ti GPUs.
Software Dependencies No The paper mentions 'Pytorch' but does not provide a specific version number. No other software components are listed with version numbers.
Experiment Setup Yes CNN parameters and one-hot parameters {wk} were optimized together with the standard Poisson loss function and Lone-hot to fit mean firing rates smoothed using a 10 ms Gaussian filter. The optimization was performed using ADAM [25]... The network was regularized with an L2 weight penalty at each layer and an L1 penalty on the model output.