Abstract: Provided are systems and methods that use a novel learning rate scheduling technique to dynamically adapt the learning rate of an adversarial model to maintain an appropriate balance between adversarial components of the model. The scheduling technique is driven by the fact that, in some settings, the loss of an ideal adversarial network can be analytically determined a priori. A scheduler component can thus operate to keep the loss of the optimized network close to that of an ideal adversarial net.

Abstract: A system including a main neural network for performing one or more machine learning tasks on a network input to generate one or more network outputs. The main neural network includes a Mixture of Experts (MoE) subnetwork that includes a plurality of expert neural networks and a gating subsystem. The gating subsystem is configured to: apply a softmax function to a set of gating parameters having learned values to generate a respective softmax score for each of one or more of the plurality of expert neural networks; determine a respective weight for each of the one or more of the plurality of expert neural networks; select a proper subset of the plurality of expert neural networks; and combine the respective expert outputs generated by the one or more expert neural networks in the proper subset to generate one or more MoE outputs.