Self-supervised back propagation for deep learning

A computer system for analyzing a first neural network via a second neural network according to a differentiable function, the second neural network comprising an objective function, the computer system comprising:

• a processor;
• and
• a memory coupled to the processor, the memory storing:
• the first neural network;
• the second neural network;
• and
• instructions that, when executed by the processor, cause the computer system to:
• add a derivative node to the first neural network;
• iteratively for n=1 to N data items in a data set, where N>2:
• compute first feed-forward activation values in the first neural network for an nth data item;
• compute derivatives of the differentiable function with respect to one or more learned parameters of each node of the first neural network;
• wherein the derivative node receives the derivative of the differentiable function with respect to the one or more learned parameters computed for a node of the first neural network;
• wherein the derivative node is connected to the second neural network such that the second neural network can receive the derivative as input;
• compute second feed-forward activation values in the second neural network for the derivative of the differentiable function provided to the second neural network from the derivative node;
• and
• compute derivatives of the objective function with respect to each of the one or more learned parameters of each node in both the first neural network and the second neural network;
• and
• after an Nth iteration, update the one or more learned parameters of the first neural network and/or the second neural network according a gradient of the objective function;
• wherein the gradient of the objective function is defined by the derivatives of the objective function for the N iterations.