Time-Varying Sequence Model

Traditional machine learning sequence models, such as RNN and LSTM, can solve sequential data problems with the use of internal memory states. However, neuron units weights are shared at each time step to reduce computational costs, limiting their ability learn time-varying relationships between model inputs outputs. In this context, paper proposes two methods characterize dynamic in real-world data, namely, (ITV model) external (ETV model). Our were designed an automated basis expansion module adapt or parameters without requiring high complexity. Extensive experiments performed on synthetic demonstrated superior prediction classification results conventional models. proposed ETV is particularly effective handling long data.