Abstract: Deep Neural Networks (DNNs) with many hidden layers and many units per layer are very flexible models with a very large number of parameters. As such, DNNs are challenging to optimize. To achieve real-time computation, embodiments disclosed herein enable fast DNN feature transformation via optimized memory bandwidth utilization. To optimize memory bandwidth utilization, a rate of accessing memory may be reduced based on a batch setting. A memory, corresponding to a selected given output neuron of a current layer of the DNN, may be updated with an incremental output value computed for the selected given output neuron as a function of input values of a selected few non-zero input neurons of a previous layer of the DNN in combination with weights between the selected few non-zero input neurons and the selected given output neuron, wherein a number of the selected few corresponds to the batch setting.

Abstract: Computing the feature Maximum Mutual Information (fMMI) method requires multiplication of vectors with a huge matrix. The huge matrix is subdivided into block sub-matrices. The sub-matrices are quantized into different values and compressed by replacing the quantized element values with 1 or 2 bit indices. Fast multiplication with those compressed matrices with far fewer multiply/accumulate operations compared to standard matrix computation is enabled and additionally obviates a de-compression method for decompressing the sub-matrices before use.

Abstract: Techniques are described for updating an automatic speech recognition (ASR) system that, prior to the update, is configured to perform ASR using a first finite-state transducer (FST) comprising a first set of paths representing recognizable speech sequences. A second FST may be accessed, comprising a second set of paths representing speech sequences to be recognized by the updated ASR system. By analyzing the second FST together with the first FST, a patch may be extracted and provided to the ASR system as an update, capable of being applied non-destructively to the first FST at the ASR system to cause the ASR system using the first FST with the patch to recognize speech using the second set of paths from the second FST. In some embodiments, the patch may be configured such that destructively applying the patch to the first FST creates a modified FST that is globally minimized.

Abstract: Techniques are described for updating an automatic speech recognition (ASR) system that, prior to the update, is configured to perform ASR using a first finite-state transducer (FST) comprising a first set of paths representing recognizable speech sequences. A second FST may be accessed, comprising a second set of paths representing speech sequences to be recognized by the updated ASR system. By analyzing the second FST together with the first FST, a patch may be extracted and provided to the ASR system as an update, capable of being applied non-destructively to the first FST at the ASR system to cause the ASR system using the first FST with the patch to recognize speech using the second set of paths from the second FST. In some embodiments, the patch may be configured such that destructively applying the patch to the first FST creates a modified FST that is globally minimized.

Abstract: Deep Neural Networks (DNNs) with many hidden layers and many units per layer are very flexible models with a very large number of parameters. As such, DNNs are challenging to optimize. To achieve real-time computation, embodiments disclosed herein enable fast DNN feature transformation via optimized memory bandwidth utilization. To optimize memory bandwidth utilization, a rate of accessing memory may be reduced based on a batch setting. A memory, corresponding to a selected given output neuron of a current layer of the DNN, may be updated with an incremental output value computed for the selected given output neuron as a function of input values of a selected few non-zero input neurons of a previous layer of the DNN in combination with weights between the selected few non-zero input neurons and the selected given output neuron, wherein a number of the selected few corresponds to the batch setting.

Abstract: Computing the feature Maximum Mutual Information (fMMI) method requires multiplication of vectors with a huge matrix. The huge matrix is subdivided into block sub-matrices. The sub-matrices are quantized into different values and compressed by replacing the quantized element values with 1 or 2 bit indices. Fast multiplication with those compressed matrices with far fewer multiply/accumulate operations compared to standard matrix computation is enabled and additionally obviates a de-compression method for decompressing the sub-matrices before use.

Abstract: A method of server site rendering 3D images on a server computer coupled to a client computer wherein the client computer instructs a server computer to load data for 3D rendering and sends a stream of rendering parameter sets to the server computer, each set of rendering parameters corresponding with an image to be rendered; next the render computer renders a stream of images corresponding to the stream of parameter sets and the stream of images is compressed with a video compression scheme and sent from the server computer to the client computer where the client computer decompresses the received compressed video stream and displays the result in a viewing port. The rendering and communication chain is subdivided in successive pipeline stages that work in parallel on successive rendered image information.

Abstract: A method of point picking on a fused volume rendered view of multiple registered volumes having the following steps: selecting a rendered pixel, finding the projection ray through the rendered pixel, finding the intersection segment of the projection ray with the fused volume, traversing the intersection segment according to a traversing intersection point, calculating the intermediate fused volume rendered value at each traversing intersection point, evaluating a stop condition at each traversing intersection point leading to a final fused volume rendered value, outputting the coordinates of the traversing intersection point where the stop condition was reached as the picked point.

Abstract: A method of server site rendering 3D images on a server computer coupled to a client computer wherein the client computer instructs a server computer to load data for 3D rendering and sends a stream of rendering parameter sets to the server computer, each set of rendering parameters corresponding with an image to be rendered; next the render computer renders a stream of images corresponding to the stream of parameter sets and the stream of images is compressed with a video compression scheme and sent from the server computer to the client computer where the client computer decompresses the received compressed video stream and displays the result in a viewing port. The rendering and communication chain is subdivided in successive pipeline stages that work in parallel on successive rendered image information.

Abstract: A method of point picking on a fused volume rendered view of multiple registered volumes having the following steps: selecting a rendered pixel, finding the projection ray through the rendered pixel, finding the intersection segment of the projection ray with the fused volume, traversing the intersection segment according to a traversing intersection point, calculating the intermediate fused volume rendered value at each traversing intersection point, evaluating a stop condition at each traversing intersection point leading to a final fused volume rendered value, outputting the coordinates of the traversing intersection point where the stop condition was reached as the picked point.