Using Fast Weights to Attend to the Recent Past
This addresses a foundational problem in neural network design for sequence modeling, offering a neurally plausible alternative to existing attention mechanisms.
The paper tackled the limitation of traditional neural networks having only two variable types by introducing 'fast weights' that change at intermediate time-scales, enabling temporary memory storage for attention to recent past sequences without storing activity copies.
Until recently, research on artificial neural networks was largely restricted to systems with only two types of variable: Neural activities that represent the current or recent input and weights that learn to capture regularities among inputs, outputs and payoffs. There is no good reason for this restriction. Synapses have dynamics at many different time-scales and this suggests that artificial neural networks might benefit from variables that change slower than activities but much faster than the standard weights. These "fast weights" can be used to store temporary memories of the recent past and they provide a neurally plausible way of implementing the type of attention to the past that has recently proved very helpful in sequence-to-sequence models. By using fast weights we can avoid the need to store copies of neural activity patterns.