The Context-Dependent Deep-Neural-Network HMM, or CD-DNN-HMM, is a recently proposed acoustic-modeling tech-nique for HMM-based speech recognition that can greatly out-perform conventional Gaussian-mixture based HMMs. For ex-ample, a CD-DNN-HMM trained on the 2000h Fisher corpus achieves 14.4 % word error rate on the Hub5’00-FSH speaker-independent phone-call transcription task, compared to 19.6% obtained by a state-of-the-art, conventional discriminatively trained GMM-based HMM. That CD-DNN-HMM, however, took 59 days to train on a modern GPGPU—the immense computational cost of the mini-batch based back-propagation (BP) training is a major road-block. Unlike the familiar Baum-Welch training for conven-tional HMMs, BP cannot be efficiently parallelized across data. In this paper we show that the pipelined approximation to BP, which parallelizes computation with respect to layers, is an efficient way of utilizing multiple GPGPU cards in a single server. Using 2 and 4 GPGPUs, we achieve a 1.9 and 3.3 times end-to-end speed-up, at parallelization efficiency of 0.95 and 0.82, respectively, at no loss of recognition accuracy. Index Terms: speech recognition, deep neural networks, paral-lelization, GPGPU

Snapping is a widely used technique that helps users position graphical objects precisely, e.g., to align them with a grid or other graphical objects. Unfortunately, whenever users want to position a dragged object close to such an aligned location, they first need to deactivate snapping. We propose snap-and-go, a snapping technique that overcomes this limitation. By merely stopping dragged objects at aligned positions, rather than "warping" them there, snap-and-go helps users align objects, yet still allows placing dragged objects anywhere else. While this approach of inserting additional motor space renders snap-and-go slightly slower than traditional snapping, snap-and-go simplifies the user interface by eliminating the need for a deactivation option and thereby allows introducing snapping to application scenarios where traditional snapping is inapplicable. In our user studies, participants were able to align objects up to 138% (1D) and 231% (2D) faster with snap-and-go than without and snap-and-go proved robust against the presence of distracting snap targets.