Machine Learning with Living Neural Networks

Biological computing architectures depend on efficient interactions between hardware and wetware systems. Understanding the information response and propagation of a living neural network is critical in building hybrid machine learning architectures. In this study, we incorporate living neural networks as an encoding layer in a hybrid variational auto-encoder. We propose that the nonlinear interactions of neural networks allow them to be used as an efficient dimension reduction layer. Additionally, we investigate how information in living neural networks is transmitted to neighboring cells by training VAEs on non-stimulated cell regions.