MRIL is a two-compartment neuron model that learns salient features within continuous information streams through self-supervising process. While MRIL is a biologically-inspired model, it is applicable to several real problems such as unsupervised spike-pattern detection or blind source separation. We have implemented neural network models for MRIL in Python, which replicate the main results in our publication, Asabuki & Fukai (2020). Any binary matrix of spike train data can be analyzed by simply attributing it to spike_mat variable in our scripts.
The code is available here.
This is a computer program that performs the blind separation of signals from a data matrix, developed for the study presented in Ghandour et al. (2019). The core algorithm is the non-negative matrix factorization (NMF) method proposed by Lee & Sheng (1999). After performing pre-processing steps on the input data, the program will automatically choose the NMF solution that minimizes the Akaike Information Criterion (AIC).
The code can be found here.
Other codes developed in the group can be found on the NCBC github page: https://github.com/oist-ncbc/