An encoder, connectable to a data-memory, for storing numerical values in the data-memory, which lie in a value range between a predefined-minimum-value and a predefined-maximum-value, the encoder including an assignment instruction, according to which the value range is subdivided into multiple discrete intervals, and the encoder being configured to classify a numerical value to be stored in exactly one interval and to output an identifier of this interval, the intervals varying in width on the scale of the numerical values. A decoder for numerical values, which are stored in a data-memory using an encoder, to assign according to one assignment instruction an identifier of a discrete interval retrieved from the data-memory a fixed numerical value belonging to this interval and to output it. Also described are an AI module including an ANN, an encoder and a decoder, and a method for manufacturing the AI module, and an associated computer program.

In artificial neural networks, and other similar applications, there is typically a large amount of data involved that is considered sparse data. Due to the large size of the data involved in such applications, it is helpful to compress the data to save bandwidth resources when transmitting the data and save memory resources when storing the data. Introduced herein is a compression technique that selects elements with significant values from data and restructures them into a structured sparse format. By generating metadata that enforces the structured sparse format and organizing the data according to the metadata, the introduced technique not only reduces the size of the data but also consistently places the data in a particular format. As such, hardware can be simplified and optimized to process the data much faster and much more efficiently than the conventional compression techniques that rely on a non-structured sparsity format.

Embodiments described herein provide a provide a fully unsupervised model for text compression. Specifically, the unsupervised model is configured to identify an optimal deletion path for each input sequence of texts (e.g., a sentence) and words from the input sequence are gradually deleted along the deletion path. To identify the optimal deletion path, the unsupervised model may adopt a pretrained bidirectional language model (BERT) to score each candidate deletion based on the average perplexity of the resulting sentence and performs a simple greedy look-ahead tree search to select the best deletion for each step.

Various embodiments may provide systems and methods for supporting lossy compression of feedback information, such as acknowledgment (ACK) information, negative acknowledgement (NACK) information, etc. Various embodiments may support lossy compression for feedback messages in retransmission systems, such as Hybrid Automatic Repeat Request (HARD) protocols, the Multiple Incremental Redundancy Scheme (MIRS), etc. Various embodiments may support compression of feedback bits using a different compression codebook per symbol or per group of symbols. Various embodiments may support selection of a compression codebook per symbol, or per group of symbols, based at least in part on a probability of successful decoding of each code block in the symbol or group of symbols.

Systems and methods for encoding and decoding neural network data is provided. A method includes: receiving a neural network representation (NNR) bitstream including a group of NNR units (GON) that represents an independent neural network with a topology, the GON including an NNR model parameter set unit, an NNR layer parameter set unit, an NNR topology unit, an NNR quantization unit, and an NNR compressed data unit; and reconstructing the independent neural network with the topology by decoding the GON.

An apparatus includes a processor and a machine-readable medium having program code to cause the apparatus to obtain a first dictionary based on a first training set of signals and determine a first subset of the first training set of signals based on a training reconstruction accuracy threshold and the first dictionary, wherein each atom in the first dictionary includes at least one of a signal pattern and a function representing the signal pattern. The program code also includes code to generate a second dictionary based on a second training set of signals, wherein the second training set of signals includes the first subset of the first training set of signals.

A computer implemented method for object detection includes: determining a grid, the grid comprising a plurality of grid cells; determining, for a plurality of time steps, for each grid cell, a plurality of respective radar detection data, each radar detection data indicating a plurality of radar properties; determining, for each time step, a respective radar map indicating a pre-determined radar map property in each grid cell; converting the respective radar detection data of the plurality of grid cells for the plurality of time steps to a point representation of pre-determined first dimensions; converting the radar maps for the plurality of time steps to a map representation of pre-determined second dimensions, wherein the pre-determined first dimensions and the pre-determined second dimensions are at least partially identical; concatenating the point representation and the map representation to obtain concatenated data; and carrying out object detection based on the concatenated data.

Methods, systems, and devices for wireless communications are described. A user equipment (UE) may perform a measurement operation to attain multiple measurements to report to a base station. The measurements may correspond to a first number of bits if reported. The UE may compress the measurements using an encoder neural network (NN) to obtain an encoder output indicating the measurements. This encoder output may include a second number of bits that is less than the first number of bits. The UE may report the encoder output to the base station in this compressed form. At the base station, the encoder output may be decompressed according to a decoder NN. Once the base station decompresses the encoder output, the UE and base station may communicate according to the measurements determined from the decompression. In some cases, the base station may perform load redistribution based on the measurements.

A system comprises an encoder configured to compress attribute information and/or spatial for a point cloud and/or a decoder configured to decompress compressed attribute and/or spatial information for the point cloud. To compress the attribute and/or spatial information, the encoder is configured to convert a point cloud into an image based representation. Also, the decoder is configured to generate a decompressed point cloud based on an image based representation of a point cloud.

A computer-implemented method of an arithmetic processing, the method including: identifying maximum absolute values of individual dimensions by projecting a maximum absolute value in a direction of each of the individual dimensions of a tensor represented by a multidimensional array, the tensor in which a value is set for each of elements of the array; identifying a minimum value that indicates a minimum maximum absolute value among the maximum absolute values of the individual dimensions; and setting a quantization range for the tensor on a basis of the minimum value.