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.

Abstract: A computer-implemented method for decreasing RAID rebuilding time may include (1) identifying data for which there is a need for physical integrity and high availability, (2) segmenting the data sequentially into a plurality of groups of chunks, with each group of chunks including redundant data sufficient to rebuild a lost chunk within the group of chunks, (3) storing the groups of chunks on a storage array according to a four-cycle-free bipartite storage map that, for each group of chunks, stores each chunk on a different device set within the storage array and, when a chunk within a group of chunks is lost, enables all other chunks within the group to be read in parallel from different devices within the storage array. Various other methods, systems, and computer-readable media are also disclosed.