Multi-ANN embedded system based on a custom 3D-DRAM

Machine Learning in the form of Artificial Neural Networks (ANNs) has gained considerable traction in applications such as image recognition and speech recognition. These applications typically employ a subset of ANNs known as Convolutional Neural Networks (CNNs) which re-use parameters and thus reduce main memory bandwidth. However, there are other types of ANN that do not provide reuse opportunities such as autoencoders and Long Short-term memory. Most research has focused on implementing CNNs but because of their extensive use of SRAM have both ANN size restrictions and performance degradation when used in applications that utilize other types of ANN. This work demon-strates how a customized 3D-DRAM with a very wide databus can be combined with application-specific layers to produce a system meeting the requirements of embedded systems employing multiple instances of disparate ANNs.