quantlab.model.predefined.rnn#

Qlib’s recurrent benchmark models: LSTM, GRU and ALSTM.

RNNNet reproduces Qlib’s LSTMModel and GRUModel (qlib/contrib/model/pytorch_lstm_ts.py, pytorch_gru_ts.py), and ALSTMNet its ALSTMModel (pytorch_alstm_ts.py, Qin et al., “A Dual-Stage Attention-Based Recurrent Neural Network for Time Series Prediction”, IJCAI 2017). LSTMRegressor, GRURegressor and ALSTMRegressor train them with Qlib’s time-series regimen (QlibSequenceHead) and its Alpha158 benchmark settings.

Classes

ALSTMNet

Qlib's ALSTMModel: a recurrent network with attention over its time steps.

ALSTMRegressor

Qlib's ALSTM benchmark model on (bar, symbol) samples.

GRURegressor

Qlib's GRU benchmark model on (bar, symbol) samples.

LSTMRegressor

Qlib's LSTM benchmark model on (bar, symbol) samples.

RNNNet

Qlib's LSTMModel / GRUModel with an output of any width.