Probabilistic Tensor Decomposition of Neural Population Spiking Activity
Authors: Hugo Soulat, Sepiedeh Keshavarzi, Troy Margrie, Maneesh Sahani
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We apply the model to neural recordings taken under conditions of visual-vestibular sensory integration, revealing how the encoding of selfand visual-motion signals is modulated by the sensory information available to the animal. Last, in section 4, we analyse synthetic data and neural recordings in mice performing passive multisensory integration [18] and show that our method can estimate the population-level effects of temporal dynamics and experimental condition in a fully probabilistic manner. We show improvements in variance explained, deviance and, last but not least, decomposition robustness when compared to standard CP and GCP baselines. |
| Researcher Affiliation | Academia | Hugo Soulat Gatsby Unit University College London London, W1T 4JG hugos@gatsby.ucl.ac.uk Sepiedeh Keshavarzi Sainsbury Wellcome Centre University College London London, W1T 4JG s.keshavarzi@ucl.ac.uk Troy W. Margrie Sainsbury Wellcome Centre University College London London, W1T 4JG t.margrie@ucl.ac.uk Maneesh Sahani Gatsby Unit University College London London, W1T 4JG maneesh@gatsby.ucl.ac.uk |
| Pseudocode | No | The paper describes the variational approach and updates but does not include any formal pseudocode or algorithm blocks. |
| Open Source Code | Yes | https://github.com/hugosou/vbgcp |
| Open Datasets | Yes | We apply the model to neural recordings taken under conditions of visual-vestibular sensory integration, revealing how the encoding of selfand visual-motion signals is modulated by the sensory information available to the animal. We then applied our method to neural spiking data [18] recorded in mice during a multisensory integration paradigm that aimed to elucidate the contribution of vestibular and visual signals to self-motion representation in the cortex. [18] Sepiedeh Keshavarzi, Edward F Bracey, Richard A Faville, Dario Campagner, Adam L Tyson, Stephen C Lenzi, Tiago Branco, and Troy W Margrie. The retrosplenial cortex combines internal and external cues to encode head velocity during navigation. bio Rxiv, page 2021.01.22.427789, jan 2021. |
| Dataset Splits | Yes | For each of the 24 cross validation folds, we split the dataset in half across trials. |
| Hardware Specification | Yes | All experiment were performed using an Intel(R) Xeon(R) CPU E5-2620 v3 @ 2.40GHz with 65GB of RAM. |
| Software Dependencies | No | The paper mentions using "standard CP and Poisson GCP" and provides a GitHub link, but it does not specify software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions, or specific library versions). |
| Experiment Setup | Yes | All algorithms were trained with a maximum of 10000 iterations. For VB-GCP, we used k0λ = 100, θ0λ = 1 (see (16)). Neuron groups (see (17)) were based on the neuron recording sites (Layer and RSP division), and the offset tensor was only allowed to vary across the neuron and experimental condition dimensions. |