Abstract: A neural processing device is provided. The neural processing device comprises: an activation buffer in which first and second input activations are stored, an activation compressor configured to generate a first compressed input activation by using the first and second input activations, and a tensor unit configured to perform two-dimensional calculations using the first compressed input activation, wherein the first compressed input activation comprises first input row data comprising at least a portion of the first input activation and at least a portion of the second input activation, and first metadata corresponding to the first input row data.

Abstract: An apparatus and method for a tensor permutation engine. The TPE may include a read address generation unit (AGU) to generate a plurality of read addresses for the plurality of tensor data elements in a first storage and a write AGU to generate a plurality of write addresses for the plurality of tensor data elements in the first storage. The TPE may include a shuffle register bank comprising a register to read tensor data elements from the plurality of read addresses generated by the read AGU, a first register bank to receive the tensor data elements, and a shift register to receive a lowest tensor data element from each bank in the first register bank, each tensor data element in the shift register to be written to a write address from the plurality of write addresses generated by the write AGU.

Abstract: The present disclosure provides a processing device including: a coarse-grained pruning unit configured to perform coarse-grained pruning on a weight of a neural network to obtain a pruned weight, an operation unit configured to train the neural network according to the pruned weight. The coarse-grained pruning unit is specifically configured to select M weights from the weights of the neural network through a sliding window, and when the M weights meet a preset condition, all or part of the M weights may be set to 0. The processing device can reduce the memory access while reducing the amount of computation, thereby obtaining an acceleration ratio and reducing energy consumption.

Abstract: Neural networks can be implemented with DNA strand displacement (DSD) circuits. The neural networks are designed and trained in silico taking into account the behavior of DSD circuits. Oligonucleotides comprising DSD circuits are synthesized and combined to form a neural network. In an implementation, the neural network may be a binary neural network in which the output from each neuron is a binary value and the weight of each neuron either maintains the incoming binary value or flips the binary value. Inputs to the neural network are one more oligonucleotides such as synthetic oligonucleotides containing digital data or natural oligonucleotides such as mRNA. Outputs from the neural networks may be oligonucleotides that are read by directly sequencing or oligonucleotides that generate signals such as by release of fluorescent reporters.

Abstract: An apparatus and method for a tensor permutation engine. The TPE may include a read address generation unit (AGU) to generate a plurality of read addresses for the plurality of tensor data elements in a first storage and a write AGU to generate a plurality of write addresses for the plurality of tensor data elements in the first storage. The TPE may include a shuffle register bank comprising a register to read tensor data elements from the plurality of read addresses generated by the read AGU, a first register bank to receive the tensor data elements, and a shift register to receive a lowest tensor data element from each bank in the first register bank, each tensor data element in the shift register to be written to a write address from the plurality of write addresses generated by the write AGU.

Abstract: The present disclosure discloses a neural network processing module, in which a mapping unit is configured to receive an input neuron and a weight, and then process the input neuron and/or the weight to obtain a processed input neuron and a processed weight; and an operation unit is configured to perform an artificial neural network operation on the processed input neuron and the processed weight. Examples of the present disclosure may reduce additional overhead of the device, reduce the amount of access, and improve efficiency of the neural network operation.

Abstract: Methods and apparatus are disclosed supporting a design flow for developing quantized neural networks. In one example of the disclosed technology, a method includes quantizing a normal-precision floating-point neural network model into a quantized format. For example, the quantized format can be a block floating-point format, where two or more elements of tensors in the neural network share a common exponent. A set of test input is applied to a normal-precision flooding point model and the corresponding quantized model and the respective output tensors are compared. Based on this comparison, hyperparameters or other attributes of the neural networks can be adjusted. Further, quantization parameters determining the widths of data and selection of shared exponents for the block floating-point format can be selected. An adjusted, quantized neural network is retrained and programmed into a hardware accelerator.

Abstract: Methods, apparatus, and computer-readable media for determining and utilizing corrections to robot actions. Some implementations are directed to updating a local features model of a robot in response to determining a human correction of an action performed by the robot. The local features model is used to determine, based on an embedding generated over a corresponding neural network model, one or more features that are most similar to the generated embedding. Updating the local features model in response to a human correction can include updating a feature embedding, of the local features model, that corresponds to the human correction. Adjustment(s) to the features model can immediately improve robot performance without necessitating retraining of the corresponding neural network model.

Abstract: Disclosed embodiments provide techniques for automated technical support based on cognitive capabilities and preferences of a user. A user profile is obtained which includes a skill level assessment. A solution path includes one or more potential solutions for a problem. One or more solutions in the solution path are presented to a user as a potential remedy for a technical problem, based on the cognitive capabilities and preferences of a user.

Abstract: An energy-efficient sequencer comprising inline multipliers and adders causes a read source that contains matching values to output an enable signal to enable a data item prior to using a multiplier to multiply the data item with a weight to obtain a product for use in a matrix-multiplication in hardware. A second enable signal causes the output to be written to the data item.

Abstract: Systems and methods for remote intervention are disclosed herein. The system can include memory including: a user profile database; a content database; and a model database. The system can include a remote device including: a network interface; and an I/O subsystem. The system can include a content management server that can: receive a first electrical signal from the remote device; generate and send an electrical signal to the remote device directing the launch of the content authoring interface; receive a second electrical signal including content received by the content authoring interface from the remote device; identify a plurality of response demands in the received content; determine a level of the received content based on the identified plurality of response demands; determine the acceptability of the received content based on the identified plurality of response demands; and generate and send an alert to the remote device.

Abstract: A method for verifying a calculation of a neuron value of multiple neurons of a neural network, including: carrying out or triggering a calculation of neuron functions of the multiple neurons, in each case to obtain a neuron value, the neuron functions being determined by individual weightings for each neuron input; calculating a first comparison value as the sum of the neuron values of the multiple neurons; carrying out or triggering a control calculation with one or multiple control neuron functions and with all neuron inputs of the multiple neurons, to obtain a second comparison value as a function of the neuron inputs of the multiple neurons and of the sum of the weightings of the multiple neurons assigned to the respective neuron input; and recognizing an error as a function of the first comparison value and of the second comparison value.

Abstract: A diffusive memristor device and an electronic device for emulating a biological neuron is disclosed. The diffusive memristor device includes a bottom electrode, a top electrode formed opposite the bottom electrode, and a dielectric layer disposed between the top electrode and the bottom electrode. The dielectric layer comprises an oxide doped with a metal.

Abstract: A system and method for evaluating the performance and usage of a cognitive computing tool which answers questions from users. A log file for these interactions includes the questions, the answers and a confidence rating assigned by the tool to each answer. Questions and answers are analyzed to determine validity, accuracy, and categories by subject matter experts or text analytics tools, and the results are added to the log file. Comments and sentiments from users may be analyzed and added to the log file. Additional data about the users, such as identities, demographics, and locations, may be added. Data from the log file may be presented in a dashboard display as metrics, such as trends and comparisons, describing the usage and performance of the cognitive computing tool. Answers may be displayed as they were presented to the users. Selectable filters may be provided to control the data displayed.

Abstract: Disclosed is a method for convolution calculation in a neural network, comprising: reading an input feature map, depthwise convolution kernels and pointwise convolution kernels from a dynamitic random access memory (DRAM); performing depthwise convolution calculations and pointwise convolution calculations according to the input feature map, the depthwise convolution kernels and the pointwise convolution kernels to obtain output feature values of a first predetermined number p of points on all pointwise convolution output channels; storing the output feature values of a first predetermined number p of points on all pointwise convolution output channels into an on-chip memory, wherein the first predetermined number p is determined according to at least one of available space in the on-chip memory, a number of the depthwise convolution calculation units, and width, height and channel dimensions of the input feature map; and repeating the above operation obtain output feature values of all points on all pointwis

Abstract: Provided are embodiments for a computer-implemented method, a system, and a computer program product for updating analog crossbar arrays. The embodiments include receiving a number used in matrix multiplication to represent using pulse generation for a crossbar array, and receiving a first bit-length to represent the number, wherein the bit-length is a modifiable bit length. The embodiments also include selecting pulse positions in a pulse sequence having the first bit length to represent the number, performing a computation using the selected pulse positions in the pulse sequence, and updating the crossbar array using the computation.

Abstract: Methods and computer systems improve a trained base deep neural network by structurally changing the base deep neural network to create an updated deep neural network, such that the updated deep neural network has no degradation in performance relative to the base deep neural network on the training data. The updated deep neural network is subsequently training. Also, an asynchronous agent for use in a machine learning system comprises a second machine learning system ML2 that is to be trained to perform some machine learning task. The asynchronous agent further comprises a learning coach LC and an optional data selector machine learning system DS. The purpose of the data selection machine learning system DS is to make the second stage machine learning system ML2 more efficient in its learning (by selecting a set of training data that is smaller but sufficient) and/or more effective (by selecting a set of training data that is focused on an important task).

Abstract: Systems and methods for deriving a concordant software neural network layer are provided. A method includes receiving first instructions configured to, using a neural network processor (NNP), process a first set of data corresponding to a neural network layer, where the NNP is configured to quantize the first set of the data to generate a set of quantized data and then perform matrix-vector multiply operations on the set of quantized data using a matrix-vector-multiplier incorporated within hardware associated with the NNP to generate a first set of results. The method further includes processing the first instructions to automatically generate second instructions configured for use with at least one processor, different from the NNP, such that the second instructions, when executed by the at least one processor to perform matrix multiply operations, generate a second set of results that are concordant with the first set of results.

Abstract: Neural network processors that have been customized based on application specific synthesis specialization parameters and related methods are described. Certain example neural network processors and methods described in the present disclosure expose several major synthesis specialization parameters that can be used for specializing a microarchitecture instance of a neural network processor to specific neural network models including: (1) aligning the native vector dimension to the parameters of the model to minimize padding and waste during model evaluation, (2) increasing lane widths to drive up intra-row-level parallelism, or (3) increasing matrix multiply tiles to exploit sub-matrix parallelism for large neural network models.

Abstract: A method, apparatus and computer program product refine an automated coding model, such as for a medical chart. For each respective candidate code from a set of candidate codes, the method predicts a probability of the respective code being contained in a medical chart. The method also selects one of the candidate codes as being contained in the medical chart based upon the probability and removes the selected candidate code from the set of candidate codes. The method then repeatedly predicts the probability of a respective code being contained in the medical chart, selects one of the candidate codes based upon the predicted probability and removes the selected candidate code from the set of candidate codes. The method further determines a categorical crossentropy loss as to permit adjustment of one or more parameters of the automated coding model.