HMM/DyNeMo: Get Inferred Parameters

In this tutorial we will get the inferred parameters from a dynamic network model (HMM/DyNeMo).

Download data and trained model

In this tutorial, we’ll download example data and a trained model from OSF. Let’s first download the (prepared) data we trained the model on.

import os

def get_data(name, rename):
    os.system(f"osf -p by2tc fetch data/{name}.zip")
    os.makedirs(rename, exist_ok=True)
    os.system(f"unzip -o {name}.zip -d {rename}")
    os.remove(f"{name}.zip")
    return f"Data downloaded to: {rename}"

# Download the dataset (approximately 162 MB)
get_data("notts_meguk_giles_prepared_5_subjects", rename="prepared_data")

Let’s also download a model. In this tutorial, we will download a trained HMM, however, this can be substituted with a DyNeMo model without any other changes being needed.

def get_model(name, rename=None):
    os.system(f"osf -p by2tc fetch trained_models/{name}.zip")
    os.makedirs(rename, exist_ok=True)
    os.system(f"unzip -o {name}.zip -d {rename}")
    os.remove(f"{name}.zip")
    return f"Model downloaded to: {rename}"

get_model("tde_hmm_notts_meguk_giles", rename="results/model")

Load Trained Model

First we need to load the trained model.

from osl_dynamics.models import load

model = load("results/model")

Get the Inferred Parameters

Latent variables

Both the HMM and DyNeMo refer to the latent variable as alpha. In the HMM, alpha corresponds to the state probability time courses, and in DyNeMo, alpha corresponds to the mixing coefficient time courses. We can get these using the get_alpha method by passing the (prepared) training data.

Let’s first load the training data.

from osl_dynamics.data import Data

data = Data("prepared_data", n_jobs=4)

Now we can get the alpha for each subject.

alpha = model.get_alpha(data)

alpha is a list of numpy arrays, one for each subject.

import pickle

os.makedirs("results/inf_params", exist_ok=True)
pickle.dump(alpha, open("results/inf_params/alp.pkl", "wb"))

Observation model

We can get the inferred state/mode means and covariances with the get_means_covariances method. Note, if we passed learn_means=False in the config, the means will be zero.

import numpy as np

means, covs = model.get_means_covariances()

np.save("results/inf_params/means.npy", means)
np.save("results/inf_params/covs.npy", covs)