Processor in Memory Cortical Extensions

Abstract: The goal of this project is to build a general and scalable engine for the acceleration of a range of machine learning and cortically inspired algorithms, including parallel deep convolutional networks, Long Short Term Memory (LSTM), Cogent Confabulation, Hierarchical Temporal Memory (HTM), Sparsey, and a spiking neuron model, Liquid State Machines. A key aspect of this the development of Processor In Memory (PiM) solutions that leverage high bandwidth 3D memories (Figure 1).