A Bayesian method for reducing bias in neural representational similarity analysis
Authors: Mingbo Cai, Nicolas W. Schuck, Jonathan W. Pillow, Yael Niv
NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To test if the proposed method indeed reduces bias, we simulated f MRI data with a predefined covariance structure and compared the structure recovered by our method with that recovered by standard RSA. Fig 2B displays the covariance matrix recovered by standard RSA (first two columns) and Bayesian RSA (last two columns), with an experiment duration of approximately 10 minutes (one run, measurement resolution: TR = 2.4 sec). We applied our method to the dataset of Connolly et al. (2012) [2]. In their experiment, participants viewed images of animals from 6 different categories during an f MRI scan and rated the similarity between animals outside the scanner. f MRI time series were pre-processed in the same way as in their work [2]. |
| Researcher Affiliation | Academia | Ming Bo Cai Princeton Neuroscience Institute Princeton University Princeton, NJ 08544 mcai@princeton.edu Nicolas W. Schuck Princeton Neuroscience Institute Princeton University Princeton, NJ 08544 nschuck@princeton.edu Jonathan W. Pillow Princeton Neuroscience Institute Princeton University Princeton, NJ 08544 pillow@princeton.edu Yael Niv Princeton Neuroscience Institute Princeton University Princeton, NJ 08544 yael@princeton.edu |
| Pseudocode | No | The paper describes the proposed method using equations and narrative text, but it does not include a dedicated pseudocode block or algorithm listing. |
| Open Source Code | Yes | Our code is freely available in Brain Imaging Analysis Kit (Brainiak) (https://github.com/Intel PNI/brainiak). |
| Open Datasets | No | The paper describes the generation of simulated data and the use of real fMRI data, but it does not specify explicit training/validation/test dataset splits in the context of machine learning model training. |
| Dataset Splits | No | The paper describes the generation of simulated data and the use of real fMRI data, but it does not specify explicit training/validation/test dataset splits in the context of machine learning model training. |
| Hardware Specification | No | The paper does not provide any specific hardware specifications (e.g., GPU/CPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions the Brain Imaging Analysis Kit (Brainiak) and AFNI, but does not provide specific version numbers for these or other software dependencies, making it difficult to fully reproduce the environment. |
| Experiment Setup | Yes | 500 voxels were simulated. For each voxel, σi was sampled uniformly from [1.0, 3.0], ρi was sampled uniformly from [ 0.2, 0.6] (our empirical investigation of example f MRI data shows that small negative autoregressive coefficient can occur in white matter), si was sampled uniformly from f [0.5, 2.0]. The average SNR was manipulated by choosing f from one of three levels {1, 2, 4} in different simulations. The duration of the experiment was manipulated by using the design matrices of run 1, runs 1-2, and runs 1-4 from one participant. Inferior temporal (IT) cortex is generally considered as the late stage of ventral pathway of the visual system, in which object identity is represented. Fig 3 shows the similarity judged by the participants and the average similarity matrix estimated from IT cortex, which shows similar structure but higher correlations between animal classes. Interestingly, the pseudo-SNR map shows that only part of the anatomically-defined ROI supports the representational structure. f MRI time series were pre-processed in the same way as in their work [2]. |