al5250/multics

Python package for multi-task compressive sensing.

MultiCS

Welcome to the Python package for multi-task compressive sensing (MultiCS)! This packages provides various algorithms to solve multiple compressive sensing tasks in parallel. This codebase accompanies the paper An Efficient Algorithm for Clustered Multi-Task Compressive Sensing by Alexander Lin and Demba Ba.

Basic Usage

The main entry point into the codebase is through the MultiTaskCompSens object. You can instantiate it as follows:

from multics.model import MultiTaskCompSens

model = MultiTaskCompSens(mode="clustered", alg="em", num_clusters=2)

There are two main parameters for this object (see our paper for more details):

• mode: This determines the type of model to use. Options include separate (i.e. not sharing any information between CS tasks), joint (i.e. sharing information between all CS tasks), and clustered (i.e. automatically learning and sharing information between clusters of tasks). If the clustered option is used, you also need to specify an additional argument num_clusters.
• alg: This determines the type of algorithm to use. Options are em (i.e. the original expectation-maximization algorithm) and cofem (i.e. the acceelerated, covariance-free version of EM proposed in our paper). If using cofem, there are also additional required parameters: num_probes and cg_tol.

After instantiating the object, the model.fit function can be used to run the inference algorithm and solve the CS tasks. For a full example of how to use this function, see the time.py script file. You can also use this script to reproduce the results in our paper.