# brainiak.eventseg package¶

Event segmentation of continuous data + event transfer between datasets.

## brainiak.eventseg.event module¶

Event segmentation using a Hidden Markov Model

Given an ROI timeseries, this class uses an annealed fitting procedure to segment the timeseries into events with stable activity patterns. After learning the signature activity pattern of each event, the model can then be applied to other datasets to identify a corresponding sequence of events.

Full details are available in: Christopher Baldassano, Janice Chen, Asieh Zadbood, Jonathan W Pillow, Uri Hasson, Kenneth A Norman Discovering event structure in continuous narrative perception and memory Neuron, Volume 95, Issue 3, 709 - 721.e5 https://doi.org/10.1016/j.neuron.2017.06.041

This class also extends the model described in the Neuron paper, by allowing transition matrices that are composed of multiple separate chains of events rather than a single linear path. This allows a model to contain patterns for multiple event sequences (e.g. narratives), and fit probabilities along each of these chains on a new, unlabeled timeseries.

class brainiak.eventseg.event.EventSegment(n_events=2, step_var=<function EventSegment._default_var_schedule>, n_iter=500, event_chains=None)

Bases: sklearn.base.BaseEstimator

Class for event segmentation of continuous fMRI data

Parameters: n_events (int) – Number of segments to learn step_var (Callable[[int], float] : default 4 * (0.98 ** (step - 1))) – The Gaussian variance to use during fitting, as a function of the number of steps. Should decrease slowly over time. n_iter (int : default 500) – Maximum number of steps to run during fitting event_chains (ndarray with length = n_events) – Array with unique value for each separate chain of events, each linked in the order they appear in the array
p_start, p_end

length n_events+1 ndarray – initial and final prior distributions over events

P

n_events+1 by n_events+1 ndarray – HMM transition matrix

ll_

ndarray with length = number of training datasets – Log-likelihood for training datasets over the course of training

segments_

list of (time by event) ndarrays – Learned (soft) segmentation for training datasets

event_var_

float – Gaussian variance at the end of learning

event_pat_

voxel by event ndarray – Learned mean patterns for each event

calc_weighted_event_var(D, weights, event_pat)

Computes normalized weighted variance around event pattern

Utility function for computing variance in a training set of weighted event examples. For each event, the sum of squared differences for all timepoints from the event pattern is computed, and then the weights specify how much each of these differences contributes to the variance (normalized by the number of voxels).

Parameters: D (timepoint by voxel ndarray) – fMRI data for which to compute event variances weights (timepoint by event ndarray) – specifies relative weights of timepoints for each event event_pat (voxel by event ndarray) – mean event patterns to compute variance around ev_var ndarray of variances for each event
find_events(testing_data, var=None, scramble=False)

Applies learned event segmentation to new testing dataset

After fitting an event segmentation using fit() or setting event patterns directly using set_event_patterns(), this function finds the same sequence of event patterns in a new testing dataset.

Parameters: testing_data (timepoint by voxel ndarray) – fMRI data to segment based on previously-learned event patterns var (float or 1D ndarray of length equal to the number of events) – default: uses variance that maximized training log-likelihood Variance of the event Gaussians. If scalar, all events are assumed to have the same variance. If fit() has not previously been run, this must be specifed (cannot be None). scramble (bool : default False) – If true, the order of the learned events are shuffled before fitting, to give a null distribution segments (time by event ndarray) – The resulting soft segmentation. segments[t,e] = probability that timepoint t is in event e test_ll (float) – Log-likelihood of model fit
fit(X, y=None)

Learn a segmentation on training data

Fits event patterns and a segmentation to training data. After running this function, the learned event patterns can be used to segment other datasets using find_events

Parameters: X (time by voxel ndarray, or a list of such ndarrays) – fMRI data to be segmented. If a list is given, then all datasets are segmented simultaneously with the same event patterns y (not used (added to comply with BaseEstimator definition)) – self the EventSegment object
model_prior(t)

Returns the prior probability of the HMM

Runs forward-backward without any data, showing the prior distribution of the model (for comparison with a posterior).

Parameters: t (int) – Number of timepoints segments (time by event ndarray) – segments[t,e] = prior probability that timepoint t is in event e test_ll (float) – Log-likelihood of model (data-independent term)
predict(X)

Applies learned event segmentation to new testing dataset

Alternative function for segmenting a new dataset after using fit() to learn a sequence of events, to comply with the sklearn Classifier interface

Parameters: X (timepoint by voxel ndarray) – fMRI data to segment based on previously-learned event patterns Event label for each timepoint
set_event_patterns(event_pat)

Set HMM event patterns manually

Rather than fitting the event patterns automatically using fit(), this function allows them to be set explicitly. They can then be used to find corresponding events in a new dataset, using find_events().

Parameters: event_pat (voxel by event ndarray) –