Dataflow matrix machines as programmable, dynamically expandable, self-referential generalized recurrent neural networks
This work addresses the need for more programmable and adaptable neural network models for researchers and developers in machine learning, though it appears incremental as it builds upon existing recurrent neural network concepts.
The paper tackles the challenge of creating more flexible and powerful neural network architectures by introducing dataflow matrix machines, which generalize recurrent neural networks to support multiple linear streams and neuron types, including higher-order neurons that dynamically update network weights and topology during runtime, enabling them to serve as a general-purpose programming platform.
Dataflow matrix machines are a powerful generalization of recurrent neural networks. They work with multiple types of linear streams and multiple types of neurons, including higher-order neurons which dynamically update the matrix describing weights and topology of the network in question while the network is running. It seems that the power of dataflow matrix machines is sufficient for them to be a convenient general purpose programming platform. This paper explores a number of useful programming idioms and constructions arising in this context.