Abstract: Methods and apparatus for processing data in a distributed inference scheme based on sparse inputs. An example method includes receiving an input at a first node. A first sparsified input is generated for a second node based on a set of features associated with the second node, which are identified based on a weight mask having non-zero values for weights associated with features upon which processing by the second node depends and zeroed values for weights associated with other features. The first sparsified input is transmitted to the second node for generating an output of the second node. A second sparsified input is received from the second node and combined into a combined input. The combined input is processed into an output of the first node. The neural network is configured to generate an inference based on processing the outputs of the first node and the second node.

Abstract: Transcoding bitrate prediction techniques can include receiving a first encoded content. A transcoder bitrate can be estimated based on regression over a video quality estimator of the first encoded content and a second encoded content. The estimated transcoder bitrate can be utilized to transcoding the first encoded content into the second encoded.

Abstract: The present disclosure relates to a method for compensating an image. The method comprises estimating transform coefficients of a frequency component for a first image based on the first image, performing a dot multiplication operation between the estimated transform coefficients and a basis function associated with the frequency component to generate a compensation image, and combining the first image and the compensation image to generate a combined image.

Abstract: Neural network (NN) model training techniques can include computing activations in a forward pass using a sparse weight matrix that is transpose invariant. The neural network (NN) model training techniques can further include computing activation gradients and weight gradients in a backward pass using the sparse weight matrix.

Abstract: An innovative low-bit-width device may include a first digital-to-analog converter (DAC), a second DAC, a plurality of non-volatile memory (NVM) weight arrays, one or more analog-to-digital converters (ADCs), and a neural circuit. The first DAC is configured to convert a digital input signal into an analog input signal. The second DAC is configured to convert a digital previous hidden state (PHS) signal into an analog PHS signal. NVM weight arrays are configured to compute vector matrix multiplication (VMM) arrays based on the analog input signal and the analog PHS signal. The NVM weight arrays are coupled to the first DAC and the second DAC. The one or more ADCs are coupled to the plurality of NVM weight arrays and are configured to convert the VMM arrays into digital VMM values. The neural circuit is configured to process the digital VMM values into a new hidden state.

Abstract: Scene aware video content encoding techniques can determine if video content is a given content type and is one of one or more given titles that include one or more given scenes. The one or more given scenes of the video content of the given type and given one of the titles can be encoded using corresponding scenes specific encoding parameter values, and the non-given scenes can be encoded using one or more general encoding parameter values. The one or more given titles can be selected based on a rate of streaming of various video content titles of the given type.

Abstract: Systems and methods for implementing data protection techniques with symbol-based variable node updates for binary low-density parity-check (LDPC) codes are described. A semiconductor memory (e.g., a NAND flash memory) may read a set of data from a set of memory cells, determine a set of data state probabilities for the set of data based on sensed threshold voltages for the set of memory cells, generate a valid codeword for the set of data using an iterative LDPC decoding with symbol-based variable node updates and the set of data state probabilities, and store the valid codeword within the semiconductor memory or transfer the valid codeword from the semiconductor memory. The iterative LDPC decoding may utilize a message passing algorithm in which outgoing messages from a plurality of multi-variable nodes are generated using incoming messages (e.g., log-likelihood ratios or L-values) from a plurality of check nodes.

Abstract: A system for pruning weights during training of a neural network includes a configurable pruning hardware unit that is configured to: receive, from a neural network training engine, inputs including the weights, gradients associated with the weights, and a prune indicator per weight; select unpruned weights for pruning; prune the unpruned weights selected for pruning; update the prune indicator per weight for the weights that are selected and pruned; and provide the updated prune indicator to the training engine for the next iteration or epoch. The pruning hardware unit can be configured to perform incremental pruning or non-incremental pruning.

Abstract: The present disclosure relates to a method and an apparatus for representation of a sparse matrix in a neural network. In some embodiments, an exemplary operation unit includes a buffer for storing a representation of a sparse matrix in a neural network, a sparse engine communicatively coupled with the buffer, and a processing array communicatively coupled with the sparse engine. The sparse engine includes circuitry to: read the representation of the sparse matrix from the buffer, the representation comprising a first level bitmap, a second level bitmap, and an element array; decompress the first level bitmap to determine whether a block of the sparse matrix comprises a non-zero element; and in response to the block comprising a non-zero element, decompress the second level bitmap using the element array to obtain the block of the sparse matrix. The processing array includes circuitry to execute the neural network with the sparse matrix.

Abstract: The present disclosure relates to a computer-implemented method for processing video data. The method comprises receiving a user input corresponding to a first picture of the video data, generating, based on the user input, prediction information of the first picture with respect a reference picture of the video data, and encoding the first picture using the prediction information.

Abstract: Methods and systems are provided for implementing preprocessing operations and augmentation operations upon image datasets transformed to frequency domain representations, including decoding images of an image dataset to generate a frequency domain representation of the image dataset; performing a resizing operation based on resizing factors on the image dataset in a frequency domain representation; performing a reshaping operation based on reshaping factors on the image dataset in a frequency domain representation; and performing a cropping operation on the image dataset in a frequency domain representation. The methods and systems may further include performing an augmentation operation on the image dataset in a frequency domain representation. Methods and systems of the present disclosure may free learning models from computational overhead caused by transforming image datasets into frequency domain representations.

Abstract: A neural network, trained on a plurality of random size data samples, can receive a plurality of inference data samples including samples of different sizes. The neural network can generate feature maps of the plurality of inference data samples. Pooling can be utilized to generate feature maps having a fixed size. The fixed size feature maps can be utilized to generate an indication of a class for each of the plurality of inference data samples.

Abstract: Vector-vector multiplication or matrix-matrix multiplication computation on computing systems can include computing a first portion of a vector-vector multiplication product based on a most-significant-bit set of a first vector and a most-significant-bit set of a second vector, and determining if the first portion of the vector-vector multiplication product is less than a threshold. If the first partial vector-vector multiplication product is not less than the threshold, a remaining portion of the vector-vector multiplication product can be computed, and a rectified linear vector-vector multiplication product can be determined for the sum of the first portion of the vector-vector multiplication product and the remaining portion of the vector-vector multiplication product.

Abstract: Disclosed are systems and methods that determine whether instances of data (e.g., forward activations, backward derivatives of activations) that are used to train deep neural networks are to be stored on-chip or off-chip. The disclosed systems and methods are also used to prune the data (discard or delete selected instances of data). A system includes a hierarchical arrangement of on-chip and off-chip memories, and also includes a hierarchical arrangement of data selector devices that are used to decide whether to discard data and where in the system the data is to be discarded.

Abstract: Systems and methods for reducing the impact of defects within a crossbar memory array when performing multiplication operations in which multiple control lines are concurrently selected are described. A group of memory cells within the crossbar memory array may be controlled by a local word line that is controlled by a local word line gating unit that may be configured to prevent the local word line from being biased to a selected word line voltage during an operation; the local word line may instead be set to a disabling voltage during the operation such that the memory cell currents through the group of memory cells are eliminated. If a defect has caused a short within one of the memory cells of the group of memory cells, then the local word line gating unit may be programmed to hold the local word line at the disabling voltage during multiplication operations.

Abstract: Methods and apparatuses for video transcoding based on spatial or temporal importance include: in response to receiving an encoded video bitstream, decoding a picture from the encoded video bitstream; determining a first level of spatial importance for a first region of a background of the picture based on an image segmentation technique; applying to the first region a first resolution-enhancement technique associated with the first level of spatial importance for increasing resolution of the first region by a scaling factor, wherein the first resolution-enhancement technique is selected from a set of resolution-enhancement techniques having different computational complexity levels; and encoding the first region using a video coding standard.

Abstract: The present disclosure relates to a method and an apparatus for map reduce. In some embodiments, an exemplary processing unit includes: a 2-dimensional (2D) processing element (PE) array comprising a plurality of PEs, each PE comprising a first input and a second input, the first inputs of the PEs in a linear array in a first dimension of the PE array being connected in series and the second inputs of the PEs in a linear array in a second dimension of the PE array being connected in parallel, each PE being configured to perform an operation on data from the first input or second input; and a plurality of reduce tree units, each reduce tree unit being coupled with the PEs in a linear array in the first dimension or the second dimension of the PE array and configured to perform a first reduction operation.

Abstract: Scene aware video content encoding techniques can determine if video content is a given content type and is one of one or more given titles that include one or more given scenes. The one or more given scenes of the video content of the given type and given one of the titles can be encoded using corresponding scenes specific encoding parameter values, and the non-given scenes can be encoded using one or more general encoding parameter values. The one or more given titles can be selected based on a rate of streaming of various video content titles of the given type.

Abstract: A method for of encoding an application screen comprises partitioning graphic data into a plurality of graphic layers and classifying each of the plurality of graphic layers as either a screen content (SC) or a non-screen content (non-SC) layer. The method further comprises classifying each of the plurality of graphic layers as either a screen content (SC) or a non-screen content (non-SC) layer. Further, the method comprises rendering and encoding the one or more SC layers using a first codec and the one or more non-SC layers using a second codec.

Abstract: The present disclosure provides methods and systems for executing a neural network. The method includes: receiving a plurality of input vectors of input data; generating, among the plurality of input vectors, an estimation value associated with a subset of an input vector based on a weight vector of the activation function; determining whether the estimation value associated with the subset of the input vector satisfies a threshold condition; and determining an output of the activation function based on the estimation value.