Long Short-Term Memory

TLDR

Recurrent backpropagation struggles to store information over long intervals due to insufficient, decaying error backflow. The authors propose the long short‑term memory (LSTM) architecture to address this limitation. LSTM employs multiplicative gate units that regulate constant error flow, is local in space and time with O complexity per step, and is evaluated on artificial data with local, distributed, real‑valued, noisy patterns. By truncating gradients only when harmless, LSTM can bridge time lags over 1000 steps, outperforms several recurrent learning methods in speed and success rate, and solves long‑lag tasks previously unsolvable by other algorithms.

Abstract

Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, LSTM can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units. Multiplicative gate units learn to open and close access to the constant error flow. LSTM is local in space and time; its computational complexity per time step and weight is O. 1. Our experiments with artificial data involve local, distributed, real-valued, and noisy pattern representations. In comparisons with real-time recurrent learning, back propagation through time, recurrent cascade correlation, Elman nets, and neural sequence chunking, LSTM leads to many more successful runs, and learns much faster. LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms.