A database management system is described that can encode data to generate a plurality of data vectors. The database management system can perform the encoding by using a dictionary. The database management system can adaptively reorder the plurality of data vectors to prepare for compression of the plurality of data vectors. During a forward pass of the adaptive reordering, most frequent values of a data vector of the plurality of data vectors can be moved-up in the data vector. During a backward pass of the adaptive reordering, content within a rest range of a plurality of rest ranges can be rearranged within the plurality of data vectors according to frequencies of the content. The reordering according to frequency can further sort the rest range by value. Related apparatuses, systems, methods, techniques, computer programmable products, computer readable media, and articles are also described.

Techniques are provided herein for more efficiently storing images that have a common subject, such as product images that share the same product in the image. Each image undergoes an adaptive tiling procedure to split the image into a plurality of tiles, with each tile identifying a region of the image having pixels with the same content. The tiles across multiple images can then be clustered together and those tiles having identical content are removed. Once all duplicate tiles have been removed from the set of all tiles across the images, the tiles are once again clustered based on their encoding scheme and certain encoding parameters. Tiles within each cluster are compressed using the best compression technique for the tiles in each corresponding cluster. By removing duplicative tile content between numerous images of the same subject, the total amount of data that needs to be stored is reduced.

A method for selectively dropping out feature elements from a tensor in a neural network is disclosed. The method includes receiving a first tensor from a first layer of a neural network. The first tensor includes multiple feature elements arranged in a first order. A compressed mask for the first tensor is obtained. The compressed mask includes single-bit mask elements respectively corresponding to the multiple feature elements of the first tensor and has a second order that is different than the first order of their corresponding feature elements in the first tensor. Feature elements from the first tensor are selectively dropped out based on the compressed mask to form a second tensor which is propagated to a second layer of the neural network.

A method for selectively dropping out feature elements from a tensor in a neural network includes receiving a first tensor from a first layer of a neural network and obtaining a compressed mask for the first tensor. N mask bits of the compressed mask are received at each of N lanes of a reconfigurable computing unit and feature elements of the first tensor are respectively received at the N lanes. Feature elements are selectively dropped out from the first tensor to generate feature elements to use as at least part of a second tensor by selecting, based on a single mask bit of the compressed mask selected based on the lane, either a zero value or a feature element received at the lane for a feature element of the second tensor. The second tensor is propagated to a second layer of the neural network.

A data processing method includes: acquiring, by one or more processors, compressed data generated from data, wherein values of the compressed data are stored at first storage locations, values of the data are stored at second storage locations; acquiring, by the one or more processors, index data includes indices indicative of the first storage locations; acquiring, by the one or more processors, at least two packed indices from the index data, the at least two packed indices being generated from the index data; and inputting, by the one or more processors, the at least two packed indices into at least two selectors.

A method for compressing a flow of CAN-bus messages, which comprises: (A) during a training stage: (a) determining at least one series-type pattern; (b) defining a compressed series-type command for each of said patterns, each command comprising parameters of: (b.1) a timestamp of a first message; (b.2) a message-ID; (b.3) a type of pattern; (b.4) an indication of a field within the messages; (b.5) a parameter value at the first message; (b.6) period between messages; and (b.7) number of messages; (B) during a compression stage: (c) dividing a record of CAN-bus messages into groups of a same message-ID; (d) within each group, finding messages of a same pattern; (e) for each series, forming a compressed command in a form as defined with values for at least several parameters; and (C) during a decompression stage: (f) using the series-type compressed commands to reconstruct the content of the series of messages.

Various embodiments include methods and devices for implementing decompression of compressed high dynamic ratio fields. Various embodiments may include receiving compressed first and second sets of data fields, decompressing the first and second compressed sets of data fields to generate first and second decompressed sets of data fields, receiving a mapping for mapping the first and second decompressed sets of data fields to a set of data units, aggregating the first and second decompressed sets of data fields using the mapping to generate a compression block comprising the set of data units.

One aspect of the present disclosure relates to a data compression method. The method includes generating, by one or more processors, compressed data from data, wherein the compressed data includes one or more unduplicated values of the data and generating, by the one or more processors, index data from the data, wherein the index data includes indices indicative of storage locations for the unduplicated values.

A data store system may include a storage device configured to store a plurality of data store tables. The data store system a further include a processor in communication with the storage device. The processor may receive a request to encode a column of a data store table from the plurality of data store tables. The processor may further generate a bit value representation of each value in the column of the data store table. The processor may further generate an index. The index may include an index value representative of each bit position of the bit value representations. The processor may further reorder bits of each bit value representation according to a predetermined pattern. The processor may further encode each reordered bit value representation according to an encoding technique. The processor may further store each encoded reordered bit value representations and the index. A method and computer-readable medium are also disclosed.

A method, computer program, and computer system is provided for point cloud coding. The method includes receiving, from a bitstream, data corresponding to a point cloud; reconstructing, based on the data, a first attribute value of a first duplicate point from among a plurality of duplicate points corresponding to a single geometry position; obtaining at least one prediction residual corresponding to at least one remaining attribute value of at least one remaining duplicate point from among the plurality of duplicate points; reconstructing the at least one remaining attribute value based on the reconstructed first attribute and the at least one prediction residual; and decoding the data corresponding to the point cloud based on the reconstructed first attribute value and the reconstructed at least one remaining attribute value.