Functional Encoding Units

About

We propose a model-based, statistical framework for clustering multiple time series that exhibit nonlinear dynamics into a-priori-unknown number of sub-groups called Functional Encoding Units (FEUs) represented in an interpretable latent space. This tool will group your population data into functional groups we call FEUs and provide explainable parameters, jump and phasicity, that describe the clusters in the FEU space.

Data

The data processed by the pipeline is two-dimensional time series data, where the first dimension represents the variables to be clustered, and the second dimension represents the corresponding time series data/timepoints. For example, consider electrophysiological data recorded from 25 neurons over a duration of 1000 seconds. Assuming that one measurement is recorded every second, the data will have a shape of (25, 1000). Each row corresponds to a different neuron, and each column represents sequential time points. This structured data should be saved in a PyTorch file with the .p extension. Conversion from other data formats to the PyTorch format is supported within our pipeline, ensuring compatibility and ease of integration for various data sources.