Abstract: Key-value (KV) cache paging schemes can improve memory management for KV caches by storing a KV cache page having key tensors and value tensors for a fixed number of tokens in a fixed-sized block in the KV cache of a worker. To further improve memory management, the schemes can be modified to implement dynamic variable quantization. Quantization level of a KV cache page can be set based on a runtime importance score of the KV cache page. In addition, the quantization level of the KV cache page can be set based on the system load. The end result is a scheme that can achieve a high compression ratio of KV cache pages in the KV cache. Fitting more KV cache pages in the KV cache can lead to higher inference throughput, higher system-level user capacity, and higher end-to-end service availability.

Abstract: A method and system for selectively encrypting and decrypting portions of a file are disclosed. The system may extract a chosen part from a document, spreadsheet, presentation, or PDF file, and replace the chosen part with user-defined text or pixels, and also encrypt the extracted selection using a user-defined password. The encrypted data may be stored in a non-viewable or non-printable area of the file, such as metadata or a Visual Basic Module. For decryption, the system may retrieve the encrypted selection, decrypt it using the password, and replace the user-defined text or pixels with the original content. The method may enable a partial encryption of sensitive information while maintaining the overall coherence and searchability of the file.

Abstract: Disclosed is a method and a device for processing a file. According to an embodiment, the method comprises selecting a portion of the file to be encrypted; encrypting the selected portion of the file as encrypted data; and amending the file by replacing the selected portion with predefined data and appending the encrypted data at an end of the file.

Abstract: A modified 2-pass version of the SoftMax operation can be implemented to address reduce computational cost without loss of accuracy, in particular for deep learning neural networks such as transformer-based neural networks and large language models (LLMs). The first pass is modified to include two scalar operations at the end. At the end of the first pass, a first scalar operation is performed to calculate a logarithm of the denominator, and a second scalar operation is performed to calculate an operand value based on a sum of the logarithm of the denominator and the maximum value. The second pass is modified to perform addition and exponentiation. In the second pass, an element of an input tensor is subtracted by the operand value to obtain an exponent, and a base is raised to the exponent. The second pass avoids divisions.

Abstract: A system and method for managing a respiratory condition of a user is disclosed. The system includes an oxygen concentrator having a compression system configured to generate oxygen enriched air for delivery to the user. A physiological sensor is configured to collect physiological data of the user. The physiological data is of one or more data types. An operational sensor is configured to collect operational data of the oxygen concentrator during operation of the oxygen concentrator. The operational data is of one or more data types. A processor is configured to receive the collected physiological data and the operational data and compute a summary parameter for values of each data type. The processor is also configured to compute a health score from the summary parameter values.

Abstract: A machine learning method comprises: receiving input data comprising a set of feature variables each indicating features of a plurality of entities; encoding partial order relations between entities of the plurality of entities as a set of logic rules; constructing a hypergraph comprising a plurality of vertices representing the entities of the plurality of entities and a plurality of hyperlinks linking similar entities of the plurality of entities; adding directed edges to the hypergraph to represent partial order interactions between entities to form a partial order hypergraph; representing the partial order hypergraph as a set of matrices; forming an objective function using the logic rules and matrices of the set of matrices; minimizing the objective function to obtain a prediction function; and outputting an indication of the prediction function.

Abstract: An oxygen concentrator (100) may have a moisture conditioning system. In some implementations, the concentrator includes a compressor to induce feed gas into the concentrator. A first pathway may receive the feed gas from the compression system. The first pathway may be configured to draw moisture to produce moisture reduced feed gas. The first pathway may lead the moisture reduced feed gas to sieve bed(s) which produce oxygen enriched air with the moisture reduced feed gas. An accumulator may be configured to receive the produced oxygen enriched air from the sieve bed(s). A second pathway from the accumulator may apply the drawn-out moisture to the produced enriched air to produce humidified enriched air. A third pathway may transfer the drawn-out moisture from the first pathway to the second pathway. An outlet coupled with the second pathway may release the humidified enriched air from the concentrator for a user.

Abstract: Methods and systems for detecting visual relations in a video are disclosed.

Abstract: Example systems, methods, and apparatus to generate optimized models for Internet of Things device are disclosed. An example apparatus includes a data receiver to collect data from a sensor of an internet of things device based a first sampling frequency and a buffer having a first buffer size; a model trainer to train a model based on the data collected from the sensor; a buffer analyzer to select a second sampling frequency and to reduce the buffer to a second buffer size, the model trainer to update the model based on the second buffer size; and a platform analyzer to: determine a duration of time that that internet of things device will take to analyze sensor data based on the updated model.

Abstract: Example systems, methods, and apparatus to generate optimized models for Internet of Things device are disclosed. An example apparatus includes a data receiver to collect data from a sensor of an internet of things device based a first sampling frequency and a buffer having a first buffer size; a model trainer to train a model based on the data collected from the sensor; a buffer analyzer to select a second sampling frequency and to reduce the buffer to a second buffer size, the model trainer to update the model based on the second buffer size; and a platform analyzer to: determine a duration of time that that internet of things device will take to analyze sensor data based on the updated model.

Abstract: A process for protecting confidentiality of audio messages that are received at an electronic computing device for providing a corresponding audio output at electronic audio output devices. In operation, the electronic computing device partitions the audio message into audio content segments, where each audio content segment is assigned to a respective confidentiality level. The electronic computing device determines a current exposure level of the electronic audio output device. The electronic computing device further determines whether the audio message is to be modified based on a current exposure level of the electronic audio output device and the respective confidentiality level assigned to each audio content segment. An audio output corresponding to the modified audio message is provided at the electronic audio output device when it is determined that the audio message is to be modified.

Abstract: An apparatus is provided to use NEMO search to train GNNs that can be used for mixed-precision quantization of DNNs. For example, the apparatus generates a plurality of GNNs. The apparatus further generates a plurality of new GNNs based on the plurality of GNNs. The apparatus also generates a sequential graph for a first DNN. The first DNN includes a sequence of quantizable operations, each of which includes quantizable parameters and is represented by a different node in the sequential graph. The apparatus inputs the sequential graph into the GNNs and new GNNs and evaluates outputs of the GNNs and new GNNs based on conflicting objectives of reducing precisions of the quantizable parameters of the first DNN. The apparatus then selects a GNN from the GNNs and new GNNs based on the evaluation. The GNN is to be used for reducing precisions of quantizable parameters of a second DNN.

Abstract: An apparatus is provided to compress DNNs using filter pruning on a per-group basis. For example, the apparatus accesses a trained DNN that includes a plurality of layers. The apparatus generates a sequential graph representation of the plurality of layers. The sequential graph representation includes a sequence of nodes. Each node is a graph representation of a layer. The apparatus clusters the layers into layer groups. A layer group includes one or more layers. The apparatus determines a pruning ratio for a layer group and prunes the filters of the layers in the layer group based on the pruning ratio. The apparatus may cluster the layers and determine the pruning ratio by using a GNN. The apparatus generates compressed layers from the layers in the layer group through the filter pruning process. The apparatus further updates the DNN by replacing the layers in the layer group with the compressed layers.

Abstract: Disclosed is a method and a device for processing a file. According to an embodiment, the method comprises selecting a portion of the file to be encrypted; encrypting the selected portion of the file as encrypted data; and amending the file by replacing the selected portion with predefined data and appending the encrypted data at an end of the file.

Abstract: Methods and systems for detecting visual relations in a video are disclosed.

Abstract: A process for protecting confidentiality of audio messages that are received at an electronic computing device for providing a corresponding audio output at electronic audio output devices. In operation, the electronic computing device partitions the audio message into audio content segments, where each audio content segment is assigned to a respective confidentiality level. The electronic computing device determines a current exposure level of the electronic audio output device. The electronic computing device further determines whether the audio message is to be modified based on a current exposure level of the electronic audio output device and the respective confidentiality level assigned to each audio content segment. An audio output corresponding to the modified audio message is provided at the electronic audio output device when it is determined that the audio message is to be modified.

Abstract: A machine learning method comprises: receiving input data comprising a set of feature variables each indicating features of a plurality of entities; encoding partial order relations between entities of the plurality of entities as a set of logic rules; constructing a hypergraph comprising a plurality of vertices representing the entities of the plurality of entities and a plurality of hyperlinks linking similar entities of the plurality of entities; adding directed edges to the hypergraph to represent partial order interactions between entities to form a partial order hypergraph; representing the partial order hypergraph as a set of matrices; forming an objective function using the logic rules and matrices of the set of matrices; minimizing the objective function to obtain a prediction function; and outputting an indication of the prediction function.

Abstract: Methods and systems for predictive analysis are disclosed. A predictive analysis method comprises receiving input data comprising an indication of a user, an indication of an item, a user feature vector indicating features of the user and an item feature vector indicating features of the item; and constructing a cross feature vector indicating values for cross features between features of the user and features of the user. Embedding vectors derived from the cross feature vector, the user feature vector and the item feature vector are input into an attention network to determine a set of attentive weights which indicate cross features between the user and item features. These cross features are used in the identification of a user item preference.

Abstract: Example systems, methods, and apparatus to generate optimized models for Internet of Things device are disclosed. An example apparatus includes a data receiver to collect data from a sensor of an internet of things device based a first sampling frequency and a buffer having a first buffer size; a model trainer to train a model based on the data collected from the sensor; a buffer analyzer to select a second sampling frequency and to reduce the buffer to a second buffer size, the model trainer to update the model based on the second buffer size; and a platform analyzer to: determine a duration of time that that internet of things device will take to analyze sensor data based on the updated model.

Abstract: A method of tracking microblog messages for relevancy to an entity identifiable by an associated text and an image is disclosed. The method comprises (i) performing a search on the microblog messages based on the associated text to obtain a first set of results; (ii) performing image detection on the first set of results based on the associated image to obtain a set of seed messages; (iii) performing a search on the microblog messages based on a set of characteristics derived from the seed messages to obtain a second set of results; and (iv) selecting entries from the first and second sets of results based on relevancy to the entity, wherein the set of characteristics are associated to the entity. A related apparatus is also disclosed.