This project includes Python utilities Jupyter notebooks for simulating the activity of hippocampal place cells as well as learning HMMs which represent these data.

In addition to the usual dependencies, i.e., numpy and scipy, you will need to install a modified version of the hmmlearn library (https://github.com/ckemere/hmmlearn.git).

The vonMisesRuns notebook creates simulated place field data for 1000 short runs through different (WorldParameters['NEnv'] = 5) circular mazes containing a large number of potential place fields (NumNeurons = 100, WorldParameters['alpha'] = Pr(active in environment) = 0.5). The data are saved in a List of Lists of np arrays (NEnvs x NSequences x NumNeurons), and then exported in a Pickle file.

The HMMForSpikingData notebook loads the Pickle file and trains HMMs using a subset of the data sequences. The remainder are used for evaluating the model log-likelihood and to generate plots.