aSTDP: A More Biologically Plausible Learning
This work addresses the challenge of making artificial neural networks more biologically plausible for researchers in computational neuroscience and neuromorphic computing, but it appears incremental as it builds on existing STDP concepts.
The paper tackles the problem of bridging biological and artificial neural network learning by introducing approximate STDP, a framework that uses only spike-timing dependent plasticity rules for supervised and unsupervised tasks, eliminating the need for global loss functions. It was verified on the MNIST dataset for classification and generation tasks, though no concrete performance numbers were provided.
Spike-timing dependent plasticity in biological neural networks has been proven to be important during biological learning process. On the other hand, artificial neural networks use a different way to learn, such as Back-Propagation or Contrastive Hebbian Learning. In this work we introduce approximate STDP, a new neural networks learning framework more similar to the biological learning process. It uses only STDP rules for supervised and unsupervised learning, every neuron distributed learn patterns and don' t need a global loss or other supervised information. We also use a numerical way to approximate the derivatives of each neuron in order to better use SDTP learning and use the derivatives to set a target for neurons to accelerate training and testing process. The framework can make predictions or generate patterns in one model without additional configuration. Finally, we verified our framework on MNIST dataset for classification and generation tasks.