NewBuffer returns an empty Buffer with the given max size.

New returns a new SingleDeepCFR policy with the given model and sample buffer.

New returns a new VRSingleDeepCFR policy with the given model and sample buffer.

ExperienceTuple is a single sample of action-result collected for training.

RegretSample is a single sample of instantaneous advantages collected for training.

ReservoirBuffer is a collection of samples held in memory.

SingleDeepCFR implements cfr.StrategyProfile, and uses function approximation to estimate strategies rather than accumulation of regrets for all infosets.

VRSingleDeepCFR implements cfr.StrategyProfile, and uses function approximation to estimate strategies rather than accumulation of regrets for all infosets.

Buffer collects samples of infoset action advantages to train a Model.

Model is a regression model that can be used to fit the given samples.

Samples must be binary marshalable, but embedding that interface breaks gob decoding.

TrainedModel is a regression model to use in DeepCFR that predicts a vector of advantages for a given InfoSet.