Abstract: A method for text compression comprises recognizing a prefix string of one or more text characters preceding a target string of a plurality of text characters to be compressed. The prefix string is provided to a natural language generation (NLG) model configured to output one or more predicted continuations each having an associated rank. If the one or more predicted continuations include a matching predicted continuation relative to the next one or more text characters of the target string, the next one or more text characters are compressed as an NLG-type compressed representation. If no predicted continuations match the next one or more text characters of the target string, a longest matching entry in a compression dictionary is identified. The next one or more text characters of the target string are compressed as a dictionary-type compressed representation that includes the dictionary index value of the longest matching entry.

Abstract: An apparatus, comprising: Mach-Zehnder modulators (MZMs) numbered MZM #1 to MZM #n that exhibit nonlinearity between an input electrical domain signal and an output optical domain signal; a waveform source for applying a first voltage waveform to a first optical arm of said MZM #1; a laser source configured to direct laser light into the two arms of at least said MZM #1 to produce a first output optical domain signal; the apparatus being provided for utilizing said first output optical signal and the nonlinearities of MZMs numbered MZM #2 to said MZM #n to produce a final output voltage waveform that, compared to said input voltage, comprises at least one of: a shorter rise time, a shorter fall time, an increased signal temporal resolution or small signal dynamic range, an increased vertical resolution for small signals or large signals, a reduced noise on small signals or large signals and an increased bandwidth.

Abstract: A system and method for deep learning using a large codeword model with homomorphically compressed and dyadically encrypted data is disclosed. The system preprocesses input data, applies homomorphic-dyadic compression and encryption, tokenizes the compressed data into sourceblocks, and assigns codewords using a codebook. These codewords are processed through a machine learning core, which can be either a conventional transformer-based architecture or a latent transformer core utilizing a variational autoencoder. The system enables secure operations on encrypted data, preserving privacy while allowing complex computations. The processed output is decrypted, decompressed, and translated to match the input modality. A neural upsampler may further enhance the output. The machine learning core is continuously trained using the processed data and additional training data, improving performance over time.

Abstract: An analog-to-digital converter can include: a charge distribution and holding module configured to sample a to-be-converted signal, and to perform subtraction on the to-be-converted signal and a target reference voltage by charge distribution, in order to generate a positive-phase output voltage and a negative-phase output voltage on a first and second electric rails, respectively; a common-mode voltage compensation module coupled with the first and second electric rails, and being configured to inject common-mode charges to compensate the distributed charges of the charge distribution and holding module, and to reduce a difference between a common-mode output voltage of the charge distribution and holding module and an expected value; and a comparator configured to provide a logic signal based on a comparison between the positive-phase output voltage and the negative-phase output voltage, where the logic signal corresponds to a target digital signal of the analog-to-digital converter.

Abstract: A DWA circuit includes: a thermometer conversion unit configured to convert an input digital signal into a thermometer code; a shift amount storage unit configured to store a shift amount; a shift unit configured to cyclically shift the thermometer code; an arrangement conversion unit configured to supply, to an analog output circuit, an output control code obtained by converting a bit arrangement of a shifted code; and an update unit configured to update the shift amount, in which the shifted code includes a plurality of unconverted bit fields, the output control code includes a plurality of converted bit fields, and the arrangement conversion unit is configured to perform arrangement conversions on a plurality of bits having a same position in a bit field in the plurality of unconverted bit fields, to arrange the plurality of bits in a same converted bit field among the plurality of converted bit fields.

Abstract: A system and methods for upsampling compressed data using a jointly trained Vector Quantized Variational Autoencoder (VQ-VAE) and neural upsampler. The system compresses input data into a discrete latent space using a VQ-VAE encoder, reconstructs the data using a VQ-VAE decoder, and enhances the reconstructed data using a neural upsampler. The VQ-VAE and neural upsampler are jointly trained using a combined loss function, enabling end-to-end optimization. The system allows for efficient compression and high-quality reconstruction of various data types, including financial time-series, images, audio, video, sensor data, and text. The learned discrete latent space can be explored and manipulated using techniques such as interpolation, extrapolation, and vector arithmetic to generate new or modified data samples. The system finds applications in data storage, transmission, analysis, and generation across multiple domains.

Abstract: An ADC includes a comparator to provide a comparator output responsive to an input voltage of the ADC and a DAC output voltage; a SAR circuit including a SAR that stores an n-bit digital code that is initialized at a beginning of a conversion phase of the ADC, where the SAR circuit is to update the digital code responsive to the comparator output, where an ADC output is responsive to the digital code at an end of the conversion phase; and a DAC to provide the DAC output voltage responsive to the digital code and a reference voltage. The DAC includes an m-bit CDAC and an (n?m)-bit RDAC to provide an intermediate voltage responsive to the n?m least-significant bits of the digital code and the reference voltage. The CDAC provides the DAC output voltage responsive to the m most-significant bits of the digital code, the intermediate voltage, and reference voltage.

Abstract: A system and method for a federated deep learning platform utilizing homomorphically-compressed and encrypted data. The system comprises multiple client devices, each with a local dataset, and a central server hosting a deep learning core. Client devices convert local data into codewords, which are also homomorphically encrypted. The central server processes these encrypted codewords without decryption, preserving data privacy. The platform supports at least two architectural variants: a conventional Transformer trained on codewords, and a Latent Transformer operating on latent space vectors. Both variants eliminate the need for embedding and positional encoding layers. The system aggregates encrypted model updates from clients, enabling collaborative learning while maintaining data confidentiality. Additional features comprise differential privacy implementation and adaptive federated optimization techniques.

Abstract: An apparatus comprising an encoder is configured to: detect a first edge in the input signal and, in response, provide a pulse generation sequence comprising the encoder being configured to: generate, in the output signal, a first pulse, wherein the first pulse is provided over first and second minimum time periods irrespective of an edge subsequent the first edge being present in the input signal; and obtain a first sample of the input signal; and obtain a second sample at an end of the first pulse; and if the first sample and the second sample are indicative of different voltage levels, generate a second pulse; or if the first and second sample and the same maintain the voltage level in the output signal.

Abstract: For compressing data, preprocessing operations are performed on raw input data. A discrete cosine transform is performed on the preprocessed data, and multiple subbands are created, where each subband represents a particular range of frequencies. The subbands are organized into multiple groups, where the multiple groups comprise a first low frequency group, a second low frequency group, and a high frequency group. A latent space representation is generated corresponding to each of the multiple groups of subbands. A first bitstream is created based on the latent space representation, and an alternate representation of the latent space is used for creating a second bitstream, enabling multiple-pass techniques for data compression. The multiple bitstreams may be multiplexed to form a combined bitstream for storage and/or transmission purposes.

Abstract: A system and methods for upsampling of decompressed genomic data after lossy compression using a neural network integrates AI-based techniques to enhance compression quality. It incorporates a novel deep-learning neural network that upsamples decompressed data to restore information lost during lossy compression, taking advantage of cross-correlations between genomic data sets.

Abstract: A system and method for compressing and restoring multi-modal data utilizing a variational autoencoder to enable homomorphic compression techniques is disclosed. Multi-modal input data, comprising at least two different data types, is compressed into a unified latent space using an encoder network of a multi-modal variational autoencoder. Homomorphic operations are performed on compressed data in the latent space. The latent space compressed data is decompressed using a decoder network of the multi-modal variational autoencoder. The system utilizes modality-specific layers and cross-modal attention mechanisms to effectively process diverse data types. The homomorphic operations enable performing computations while the data is in a compressed form, preserving results of those operations in the decompressed output. This approach allows for efficient storage, transmission, and analysis of multi-modal data while maintaining privacy and data integrity across different modalities.

Abstract: A measurement unit is disclosed and includes a converter unit and a processing unit is configured to provide a measurement result value, based on a first input signal and a second input signal. The converter unit is configured to provide a first digital, quantized values based on the first input signal or derived from the first input signal and the second input signal. The converter unit is further configured to provide second digital, quantized values based on the second input signal. The measurement unit is configured to change the one or more control signals of the converter unit between determination of different first values or a determination of the different second values, wherein different first values and/or different second values are provided using different converter quantization step sizes.

Abstract: A system and method for data compression with protocol adaptation, that utilizes a codebook generator which leverages one or more machine/deep learning algorithms trained on at least a plurality of protocol policies in order to generate a protocol appendix and codebook, wherein original data is encoded by an encoder according to the codebook and sent to a decoder, but instead of just decoding the data according to the codebook to reconstruct the original data, data manipulation rules such as mapping and transformation are applied at the decoding stage to transform the decoded data into protocol formatted data.

Abstract: A system and method for event-driven data communication using codebooks with protocol prediction and translation. This invention presents an advanced adaptive communication system that dynamically optimizes network protocols using machine learning-driven prediction and translation modules. The system analyzes real-time traffic patterns and historical data to anticipate communication needs, proactively switching to optimal protocols when beneficial. A sophisticated translation module, powered by large language models, enables seamless communication between systems using different protocols, including legacy systems. This approach enhances network efficiency, ensures backward compatibility, and future-proofs communication infrastructures, making it particularly valuable in complex, heterogeneous network environments.

Abstract: Digital-to-analog converter circuitry includes sequence of multiple current drive modules. The sequence may include a first current drive module and a second current drive module of a digital-to-analog converter. The first current drive module is switchable between: i) a first mode of producing a first reference current that is mirrored by a second current drive module coupled to the first current drive module; and ii) a second mode of mirroring a second reference current that is produced by the second current drive module or a third current drive module coupled to the first current drive module.

Abstract: A switched capacitor circuit, including a metal-oxide-semiconductor field-effect transistor-based switch comprising: a first metal-oxide-semiconductor field-effect transistor having a gate, a source and a drain, wherein the source is connected to a first node and the drain is connected to a second node or wherein the drain is connected to the first node and the source is connected to the second node; a second metal-oxide-semiconductor field-effect transistor having a gate, a source and a drain, wherein the source is connected to the drain and the source and the drain are together connected to the second node; a first capacitor connected between the first node and a third node; and a second capacitor connected between the second node and the third node.

Abstract: In described examples, a circuit includes a multiplexer. The multiplexer receives an input voltage and a calibration signal. An analog-to-digital converter (ADC) is coupled to the multiplexer and generates an output code in response to the calibration signal. A storage circuit is coupled to the ADC and stores the input code representative of the calibration signal at an address corresponding to the output code. The stored input code includes an index value and a coarse value.

Abstract: The linear measuring system includes one part of a (usually steel) rod, beam or strip that is profiled with a trapezium like (subset of a) De Bruijn sequence. Each element of the alphabet is represented by a discrete height of the profile. The profile can be coated with a material with low magnetic susceptibility to enable the application on a piston rod of a pressure medium cylinder. The second part of the measuring system reads the subsequences of the main sequence by using the signals of arrays of linear Hall effect sensors, placed in a magnetic field as input for a signal processing algorithm. With a look-up table, the position of this second part of the measuring system with respect to the profile is determined. An interpolation algorithm can be used to improve the resolution to well below the sequence pitch.

Abstract: A tactile reproduction system is made more efficient by achieving a data amount reduction of a tactile signal while ensuring reproducibility of a tactile sense. A decoding apparatus according to the present technology includes a decoding unit configured to decode tactile coded data obtained by performing encoding of compressing an information amount, on a tactile signal using higher-order perception in a tactile sense. Therefore, a data amount reduction of a tactile signal can be performed in accordance with a tactile characteristic of a human.