Abstract: Functions are added to a deep neural network (“DNN”) computation graph for encoding data structures during a forward training pass of the DNN and decoding previously-encoded data structures during a backward training pass of the DNN. The functions added to the DNN computation graph can be selected based upon on the specific layer pairs specified in the DNN computation graph. Once a modified DNN computation graph has been generated, the DNN can be trained using the modified DNN computation graph. The functions added to the modified DNN computation graph can reduce the utilization of memory during training of the DNN.

Abstract: Techniques for compressing a neural network model by mixing compression ratios (sparsity patterns) are described. The weight tensor of a neural network model is divided into weight groups. The pruning cost of compressing the weight values according to a compression ratio is determined for each weight group, and a pruning cost distribution for the compression ratio is generated from the pruning costs of the weight groups. A cost threshold can then be selected from the pruning cost distribution, and weight groups having a pruning cost below the selected cost threshold are compressed according to the compression ratio. The remaining weight groups can be compressed using one or more less aggressive compression ratios. The cost threshold can be adjusted to tune the overall sparsity and accuracy of the compressed neural network.

Abstract: Functions are added to a deep neural network (“DNN”) computation graph for encoding data structures during a forward training pass of the DNN and decoding previously-encoded data structures during a backward training pass of the DNN. The functions added to the DNN computation graph can be selected based upon on the specific layer pairs specified in the DNN computation graph. Once a modified DNN computation graph has been generated, the DNN can be trained using the modified DNN computation graph. The functions added to the modified DNN computation graph can reduce the utilization of memory during training of the DNN.

Abstract: This document relates to trend response management. One example can detect a trending topic and identify content associated with the trending topic. The example can take an action relating to the content that decreases a trend-induced processing spike and/or enhances a user search experience associated with the trending topic.

Abstract: This document relates to trend response management. One example can detect a trending topic and identify content associated with the trending topic. The example can take an action relating to the content that decreases a trend-induced processing spike and/or enhances a user search experience associated with the trending topic.

Abstract: An illustrative embodiment provides a computer-implemented process for managing speculative assist threads for data pre-fetching that analyzes collected source code and cache profiling information to identify a code region containing a delinquent load instruction and generates an assist thread, including a value for a local version number, at a program entry point within the identified code region. Upon activation of the assist thread the local version number of the assist thread is compared to the global unique version number of the main thread for the identified code region and an iteration distance between the assist thread relative to the main thread is compared to a predefined value. The assist thread is executed when the local version number of the assist thread matches the global unique version number of the main thread, and the iteration distance between the assist thread relative to the main thread is within a predefined range of values.