Methods and systems using improved training and learning for deep neural networks

A system for a server comprising:

• a processing core including a deep neural network (DNN) having a plurality of L layers, each L layer includes a plurality of nodes, for each L layer in the plurality of layers, nodes are randomly connected to nodes of a L+1 layer, wherein L an integer, and for each L+1 layer in the plurality of L layers, nodes of each L+1 layer are connected to nodes of a subsequent L layer in a one-to-one manner;
• an I/O controller hub coupled to the processing core to provide network, data storage, and DNN access;
• and
• a graphics processor to fix parameters related to the nodes of each L layer and to update parameters related to the nodes of each L+1 layers.

A system for a server comprising:

• a processing core;
• an I/O controller hub coupled to the processing core to provide network, data storage, and access to a deep convolutional neural network (CNN);
• and
• a graphics processor to calculate tensors for nodes of an input layer, output layer, and a plurality of hidden layers of the CNN and to perform a deep convolutional Gaussian regression process to estimate similarity between different pairs of inputs and labels between a first sample and a second sample using the calculated tensors.