English

Inference with Hybrid Bio-hardware Neural Networks

Neural and Evolutionary Computing 2019-09-06 v2 Machine Learning Neurons and Cognition

Abstract

To understand the learning process in brains, biologically plausible algorithms have been explored by modeling the detailed neuron properties and dynamics. On the other hand, simplified multi-layer models of neural networks have shown great success on computational tasks such as image classification and speech recognition. However, the computational models that can achieve good accuracy for these learning applications are very different from the bio-plausible models. This paper studies whether a bio-plausible model of a in vitro living neural network can be used to perform machine learning tasks and achieve good inference accuracy. A novel two-layer bio-hardware hybrid neural network is proposed. The biological layer faithfully models variations of synapses, neurons, and network sparsity in in vitro living neural networks. The hardware layer is a computational fully-connected layer that tunes parameters to optimize for accuracy. Several techniques are proposed to improve the inference accuracy of the proposed hybrid neural network. For instance, an adaptive pre-processing technique helps the proposed neural network to achieve good learning accuracy for different living neural network sparsity. The proposed hybrid neural network with realistic neuron parameters and variations achieves a 98.3% testing accuracy for the handwritten digit recognition task on the full MNIST dataset.

Keywords

Cite

@article{arxiv.1905.11594,
  title  = {Inference with Hybrid Bio-hardware Neural Networks},
  author = {Yuan Zeng and Zubayer Ibne Ferdous and Weixiang Zhang and Mufan Xu and Anlan Yu and Drew Patel and Xiaochen Guo and Yevgeny Berdichevsky and Zhiyuan Yan},
  journal= {arXiv preprint arXiv:1905.11594},
  year   = {2019}
}