A method for adapting a trained neural network is provided. Input data is input to the trained neural network and a plurality of filters are applied to generate a plurality of channels of activation data. Differences between corresponding activation values in the plurality of channels of activation data are calculated and an order of the plurality of channels is determined based on the calculated differences. The neural network is adapted so that it will output channels of activation data in the determined order. The ordering of the channels of activation data is subsequently used to compress activation data values by taking advantage of a correlation between activation data values in adjacent channels.

Disclosed are a data compression method, a computer-readable storage medium, and an electronic device. The method includes: converting each data in a to-be-compressed data set into binary data in a preset format; determining a to-be-compressed bit and a significant bit for the each data in the to-be-compressed data set based on a sequence of all bits of the binary data; determining a compression bit width corresponding to the to-be-compressed data set based on bit widths of the significant bits; compressing the each data in the to-be-compressed data set based on the compression bit width, to obtain a compressed data set; and generating attribute information of the compressed data set. According to the present disclosure, the significant bit can be determined based on the sequence of all bits without adjusting orders of the bits of the binary data, thereby simplifying a data compression process and improving efficiency of data compression.

Technologies for providing shared memory for accelerator sleds includes an accelerator sled to receive, with a memory controller, a memory access request from an accelerator device to access a region of memory. The request is to identify the region of memory with a logical address. Additionally, the accelerator sled is to determine from a map of logical addresses and associated physical address, the physical address associated with the region of memory. In addition, the accelerator sled is to route the memory access request to a memory device associated with the determined physical address.

There is provided a computer implemented method of compressing a baseline dataset comprising a sequence of a plurality of instances of a plurality of unique data elements, the method comprising: providing a weight function that calculates an increasing value for a weight for each one of the plurality of instances of each one of the plurality of unique data elements in the baseline dataset, as a function of increasing number of previously processed sequential locations of each of the plurality of instances of each respective unique data element within the baseline dataset relative to a current sequential location of the baseline dataset, computing an encoding for the baseline dataset according to a distribution of the weight function computed for the plurality of unique data elements in the baseline dataset, and creating a compressed dataset according to the encoding.

A method of encoding input data includes dividing the input data into a plurality of data packets, an input packet of the plurality of data packets including a plurality of digits in a first base system, base-converting the input packet from the first base system to generate a base-converted packet including a plurality of converted digits in a second base system, the second base system having a base value lower than that of the first base system, and incrementing the converted digits to generate a coded packet for transmission through a communication channel.

A system and method for compacting data that uses mismatch probability estimation to improve entropy encoding methods to account for, and efficiently handle, previously-unseen data in data to be compacted. Training data sets are analyzed to determine the frequency of occurrence of each sourceblock in the training data sets. A mismatch probability estimate is calculated comprising an estimated frequency at which any given data sourceblock received during encoding will not have a codeword in the codebook. Entropy encoding is used to generate codebooks comprising codewords for data sourceblocks based on the frequency of occurrence of each sourceblock. A “mismatch codeword” is inserted into the codebook based on the mismatch probability estimate to represent those cases when a block of data to be encoded does not have a codeword in the codebook. During encoding, if a mismatch occurs, a secondary encoding process is used to encode the mismatched sourceblock.

In accordance with an embodiment, the method includes determining a second sequence of numbers of digits for encoding the respective integer coefficient values of the first sequence, the second sequence including, as first element, a first number of digits for encoding the first integer coefficient value of the first sequence, and as second and subsequent elements, constrained numbers of digits that are greater than or equal to respective minimum required numbers of digits for encoding the second and subsequent integer coefficient values of the first sequence. The constrained numbers of digits are such that any two successive elements of the second sequence do not differ from each other by more than a given threshold value. The method further includes encoding difference values between the successive elements of the second sequence; and encoding the integer coefficient values of the first sequence using the respective numbers of digits of the second sequence.

A method and apparatus for variable rate compression with a conditional autoencoder is herein provided. According to one embodiment, a method for compression includes receiving a first image and a first scheme as inputs for an autoencoder network; determining a first Lagrange multiplier based on the first scheme; and using the first image and the first Lagrange multiplier as inputs, computing a second image from the autoencoder network. The autoencoder network is trained using a plurality of Lagrange multipliers and a second image as training inputs.

A data compressor a zero-value remover, a zero bit mask generator, a non-zero values packer, and a row-pointer generator. The zero-value remover receives 2.sup.N bit streams of values and outputs 2.sup.N non-zero-value bit streams having zero values removed from each respective bit stream. The zero bit mask generator receives the 2.sup.N bit streams of values and generates a zero bit mask for a predetermined number of values of each bit stream in which each zero bit mask indicates a location of a zero value in the predetermined number of values corresponding to the zero bit mask. The non-zero values packer receives the 2.sup.N non-zero-value bit streams and forms a group of packed non-zero values. The row-pointer generator that generates a row-pointer for each group of packed non-zero values.

A method is disclosed to reduce computation in a self-attention deep-learning model. A feature-map regularization term is added to a loss function while training the self-attention model. At least one low-magnitude feature is removed from at least one feature map of the self-attention model during inference. Weights of the self-attention model are quantized after the self-attention model has been trained. Adding the feature-map regularization term reduces activation values of feature maps, and removing the at least one low-magnitude feature from at least one feature map may be performed by setting the low-magnitude feature to be equal to zero based on the low-magnitude feature having a value that is less than a predetermined threshold. Feature maps of the self-attention model quantized and compressed.