Rotation-invariant clustering of neuronal responses in primary visual cortex
Authors: Ivan Ustyuzhaninov, Santiago A. Cadena, Emmanouil Froudarakis, Paul G. Fahey, Edgar Y. Walker, Erick Cobos, Jacob Reimer, Fabian H. Sinz, Andreas S. Tolias, Matthias Bethge, Alexander S. Ecker
ICLR 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We apply this method to a dataset of 6000 neurons and visualize the preferred stimuli of the resulting clusters. Our results highlight the range of non-linear computations in mouse V1. We apply our method to the published model and data of Ecker et al. (2019) that contains recordings of around 6000 neurons in mouse V1 under stimulation with natural images. We start by demonstrating on a synthetic dataset (Sec. 4) that optimising Eq. (5) can successfully align the readouts (Fig. 4A). We next ask whether the alignment procedure still works in the presence of observational noise (Fig. 4B). We evaluate the GMM used to cluster R for different numbers of clusters. The test likelihood starts to plateau at around 100 clusters (Fig. 5), so we use 100 clusters in the following. |
| Researcher Affiliation | Academia | 1 Centre for Integrative Neuroscience, University of Tübingen, Germany 2 Bernstein Center for Computational Neuroscience, University of Tübingen, Germany 3 Institute for Theoretical Physics, University of Tübingen, Germany 4 Department of Neuroscience, Baylor College of Medicine, Houston, TX, USA 5 Center for Neuroscience and Artificial Intelligence, BCM, Houston, TX, USA 6 Department of Electrical and Computer Engineering, Rice University, Houston, TX, USA |
| Pseudocode | No | The paper does not contain any pseudocode or clearly labeled algorithm blocks. |
| Open Source Code | No | The paper mentions using "the pre-trained model provided by Ecker et al. (2019)" but does not state that the code for the methodology described in *this* paper is open-source or provide a link. |
| Open Datasets | Yes | We use the same dataset as in Ecker et al. (2019), consisting of simultaneous recordings of responses of 6005 excitatory neurons in mouse primary visual cortex (layers 2/3 and 4). |
| Dataset Splits | No | To obtain a quantitative estimate of the number of clusters in R, we randomly split the dataset of 6005 neurons into training (4000 neurons) and test (2005 neurons) sets, fit GMMs with different numbers of clusters on the training set, and then evaluate the likelihood of the fitted model on the test set. The paper only specifies training and test sets, without explicitly defining a separate validation set or its size. |
| Hardware Specification | No | The paper does not specify any hardware details such as GPU models, CPU types, or memory used for running the experiments. |
| Software Dependencies | No | The paper mentions "Gaussian mixture model implemented in scikit-learn (Pedregosa et al., 2011)" and "Adam (Kingma & Ba, 2015)", but it does not provide specific version numbers for these software components or any other libraries. |
| Experiment Setup | Yes | We fit models for 20 log-spaced values of β in [0.001, 10], and choose for analysis the one with the smallest alignment loss (Eq. (3)) among the models with optimised temperature T > 5. We use Adam (Kingma & Ba, 2015) with early stopping and initial learning rate of 0.01 decreased three times. |