DeepSITH: Efficient Learning via Decomposition of What and When Across Time Scales
This addresses the problem of efficient temporal learning in machine learning, particularly for tasks requiring multi-scale time processing, and appears to be a novel method rather than incremental.
The paper tackles the challenge of learning temporal relationships across multiple time scales in neural networks, which is critical for real-world applications, by introducing DeepSITH, a network with biologically-inspired SITH modules that achieve state-of-the-art performance on time series prediction and decoding tasks.
Extracting temporal relationships over a range of scales is a hallmark of human perception and cognition -- and thus it is a critical feature of machine learning applied to real-world problems. Neural networks are either plagued by the exploding/vanishing gradient problem in recurrent neural networks (RNNs) or must adjust their parameters to learn the relevant time scales (e.g., in LSTMs). This paper introduces DeepSITH, a network comprising biologically-inspired Scale-Invariant Temporal History (SITH) modules in series with dense connections between layers. SITH modules respond to their inputs with a geometrically-spaced set of time constants, enabling the DeepSITH network to learn problems along a continuum of time-scales. We compare DeepSITH to LSTMs and other recent RNNs on several time series prediction and decoding tasks. DeepSITH achieves state-of-the-art performance on these problems.