modules.higher_order.Predictor

Predictor Objects#

class Predictor(Experiment)

A class providing methods for the computation of (hard or soft) cluster \ assignments for a stream of patients from a precomputed model. Note that we don't require identical set of variables for the precomputed \ model and the predicted assignments, \ we use the intersection of the two sets for prediction. Currently, two modes for cluster prediction are provided: \ batch (called do as this is the default) and iterative \ (called do_iterative). The iterative predictor needs preprocessing in order to achieve \ scalable prediction, i.e., one that depends only on the number of nonzero coordinates in the input instance.

do#

| do(model, data)

Predicts (hard and soft) cluster assignments using a batch or \ iterative (one example at time ) mode @param cluster_model: a precomputed model containing the M and \ omega matrices for centers and prior cluster weight \ distributions @param data: a stream of patients

do_iterative#

| do_iterative(model, data)

Predicts (hard and soft) cluster assignments in an iterative mode. It uses preprocessing for fast computation, more experments needed \ to establish if the approach is more officient than the batch mode

do_batch#

| do_batch(model, data)

Predicts (hard and soft) cluster assignments @param cluster_model: a precomputed model containing the M and \ omega matrices for centers and prior cluster weight \ distributions @param data: a stream of patients