Dynamic Prediction Length for Time Series with Sequence to Sequence Networks
This addresses a limitation in time series prediction for applications like finance, but it is incremental as it builds on existing sequence-to-sequence frameworks.
The paper tackles the problem of requiring predetermined output lengths in sequence-to-sequence models for time series by developing a model that predicts variable-length outputs, using a new loss function to balance accuracy and length. It shows that the model makes longer predictions for more stable securities and effectively balances these factors, though no concrete numbers are provided.
Recurrent neural networks and sequence to sequence models require a predetermined length for prediction output length. Our model addresses this by allowing the network to predict a variable length output in inference. A new loss function with a tailored gradient computation is developed that trades off prediction accuracy and output length. The model utilizes a function to determine whether a particular output at a time should be evaluated or not given a predetermined threshold. We evaluate the model on the problem of predicting the prices of securities. We find that the model makes longer predictions for more stable securities and it naturally balances prediction accuracy and length.