Estimating Nonlinear Neural Response Functions using GP Priors and Kronecker Methods
Authors: Cristina Savin, Gasper Tkacik
NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Using artificial data we show that inference and learning in our model can robustly recover the underlying structure of neural responses even in the experimentally realistic setting where the sampling of the input space is sparse and strongly non-uniform (due to stereotyped animal behavior). We further argue for the utility of spectral mixture kernels as a powerful tool for detecting complex functional relationships beyond simple smoothing/interpolation. We go beyond artificial data that follows the assumptions of the model exactly, and show robust estimation of tuning properties in several experimental recordings. |
| Researcher Affiliation | Academia | Cristina Savin IST Austria Klosterneuburg, AT 3400 csavin@ist.ac.at Gasper Tkaˇcik IST Austria Klosterneuburg, AT 3400 tkacik@ist.ac.at |
| Pseudocode | No | The paper describes algorithmic steps but does not include any explicitly labeled "Pseudocode" or "Algorithm" blocks in the main text. |
| Open Source Code | Yes | Our implementation is based on the gpml library [9] and the code is available online. |
| Open Datasets | No | The paper mentions using "CA1 data" and thanks "Jozsef Csicsvari for kindly sharing the CA1 data", but it does not provide concrete access information (link, DOI, repository, or a formal citation to a publicly available dataset). |
| Dataset Splits | No | The paper describes using "subsets of the data constructed by combining every 5th data point" and "6min subsets" for robustness checks and validation of their estimates, but it does not specify formal training, validation, and test dataset splits needed to reproduce the experimental setup for model training and evaluation. |
| Hardware Specification | No | The paper states 'it takes minutes on a laptop to estimate a 3D field for a 30min dataset', but it does not provide specific hardware details such as CPU/GPU models, processor types, or memory specifications used for the experiments. |
| Software Dependencies | No | The paper states 'Our implementation is based on the gpml library [9]' but does not provide specific version numbers for this library or any other software dependencies. |
| Experiment Setup | Yes | The spike counts are measured within a time window δt for which the input is roughly constant (25.6ms, given by the frequency of positional tracking). Positional information was discretized on a 32 × 32 grid, corresponding to a spacing of 2.5cm. We calibrated diffusion parameters to roughly match CA1 statistics (average speed 5cm/sec, peak firing 5-10Hz, 10-30min long sessions). |