Patents.

Systems and methods for anonymizing private data for use in machine learning models

Described herein are techniques for performing a task using a trained external machine learning model. A user request comprising a task and one or more protected health information (PHI) parameters associated with a patient may be received. Using trained machine learning models, the PHI parameters may be extracted and a revised user request generated by replacing the PHI parameters with synthetic PHI parameters, which is then provided to the trained external machine learning model.

Systems and methods for modeling distributed service systems by a ledger system

A method and apparatus are described for modeling a software system of a distributed services system. The method includes receiving a stream of ledger transaction messages generated by a software system, which are then stored in a data store. A ledger system accesses a data model representing a state machine of the proper functioning of the software system, and a ledger clearance operation is performed using the state machine and the data stored within the stream of ledger transaction messages.

Magnitude Invariant Multimodal Agent for Efficient Image-Text Interface Automation

A multimodal agent architecture for automating image-text interfaces whose behaviour is invariant to input magnitude/resolution, enabling efficient UI automation by a vision-language model.

Digital asset custodial system

A digital asset custodial system for maintaining custody of, and controlling access to, cryptocurrencies and/or other digital assets, is disclosed.

Cloud storage for database using object storage

Systems and methods for text-to-image generation using language models

Embodiments described herein provide a mechanism for replacing existing text encoders in text-to-image generation models with more powerful pre-trained language models. Specifically, a translation network is trained to map features from the pre-trained language model output into the space of the target text encoder. The training preserves the rich structure of the pre-trained language model while allowing it to operate within the text-to-image generation model. The resulting modularized text-to-image model receives prompt and generates an image representing the features contained in the prompt.

Data lifecycle discovery and management (continuation)

Secure storage encryption system

Dynamic data center equipment analysis for service level agreement recommendation

Using a failure probability model, a probability of a failure in a first system within a specified time period is predicted. Using an impact model, an impact of the failure on a second system is predicted. Using a resilience model, an impact reduction of the failure is predicted. The probability, the impact, and the impact reduction are mapped into a recommended service level agreement category for the first system, the mapping performed using a classifier model.

Display screen or portion thereof with animated graphical user interface

Design patent. The ornamental design for a display screen or portion thereof with animated graphical user interface as shown and described.

Automatically classifying heterogenous documents using machine learning techniques

Methods, systems, and computer program products for automatically classifying heterogeneous documents using machine learning techniques are provided herein. A computer-implemented method includes extracting multiple features from one or more sets of heterogeneous documents; arranging at least a portion of the extracted features into multiple hierarchically-defined classes; selecting one or more of the documents associated with each of the multiple hierarchically-defined classes; training at least one machine learning model using at least a portion of the selected documents; and automatically classifying at least one input document into one or more of the multiple hierarchically-defined classes by processing one or more portions of the at least one input document using the at least one trained machine learning model.

Efficient extraction of provenance information from database query execution logs

Systems, methods, and devices are described for efficient extraction of provenance information from database query execution logs. A set of event records extracted from one or more database query execution logs is received. Groups of related event records are identified, and each defined as an activity. For each activity, an activity data structure is generated based on the associated event records. The data structure encodes query execution dependencies for the activity. For each activity, runtime information is extracted by identifying and instantiating process entities and relationships based on the associated event records and activity data structure. For each activity, lineage information is extracted by identifying and instantiating dataset entities and their lineage relationships based on query text within the associated event records. The lineage information is mapped to the runtime information for each activity to generate and store a provenance data model for use by one or more provenance applications.

Systems and methods for multimodal layout designs of digital publications

Embodiments described herein provide systems and methods for multimodal layout generations for digital publications. The system may receive as inputs, a background image, one or more foreground texts, and one or more foreground images. Feature representations of the background image may be generated. The foreground inputs may be input to a layout generator which has cross attention to the background image feature representations in order to generate a layout comprising of bounding box parameters for each input item. A composite layout may be generated based on the inputs and generated bounding boxes. The resulting composite layout may then be displayed on a user interface.

Systems and methods for controllable image generation

Embodiments described herein provide a method of image generation. The method includes a fixed diffusion model, and a trainable diffusion model. The fixed diffusion model may be pretrained on a large training corpus. The trainable diffusion model may be used to control the image generation of the fixed diffusion model by modifying internal representations of the fixed diffusion model. A task instruction may be provided in addition to a text prompt, and the task instruction may guide the trainable diffusion model together with the visual conditions. The visual conditions may be adapted according to the task instruction. During training, a fixed number of task instructions may be used. At inference, unseen task instructions may be used by combining convolutional kernels of the visual condition adapter.

Feature selection in vertical federated learning

Systems and techniques that facilitate feature selection in vertical federated learning are provided. For example, one or more embodiments described herein can comprise a system, which can comprise a memory that can store computer executable components. The system can also comprise a processor, operably coupled to the memory that can execute the computer executable components stored in memory. The computer executable components can comprise an aggregator machine learning model that aggregates a plurality of embedding components from one or more local machine learning models and removes one or more embedding components based on minimizing weights at an input layer of the aggregator machine learning model.

Explainability-driven control of an alert suppression policy

An approach is provided for controlling an alert suppression policy. Using historical training data and for a time period, an expected alert reduction for an alert suppression policy (ASP) is computed. The ASP is recommended for an information technology (IT) environment. During runtime and for the time period, an observed alert reduction for the ASP is computed. A visualization of a benefit of the ASP is generated by generating a visualization of a comparison of the observed alert reduction and the expected alert reduction. An explanation of why the ASP is recommended for the IT environment is generated by determining persistent regions for respective event time series from the historical training data and generating a visualization of the persistent regions on a time graph that includes a duration and a number of events for a given persistent region. Each event time series corresponds to a given resource for the ASP.

Domain-specific system and method for enhancing firmographic search through query understanding and expansion

Domain-specific system and method for enhancing firmographic search through query understanding and expansion presents an innovative approach to revolutionizing firmographic search processes. This system employs advanced domain-specific query understanding and expansion techniques to bridge the gap between natural language queries and structured firmographic data, significantly improving precision and relevance in search results. This method utilizes a multi-model approach to dissect natural language queries into named entities, subsequently mapping them to specific structured attributes within a database. This refined query understanding enhances the alignment of unstructured queries with the structured format required for accurate data retrieval. The system also introduces custom confidence scoring to assess the reliability of model outputs, further improving the accuracy of structured query formulation.

Byzantine-resilient bilevel federated learning

An intermediate global lower-level machine learning model is generated by executing federated learning model training using data from an identified first set of lower-level non-Byzantine client computers which are characterized as being non-Byzantine for the lower-level system. A global lower-level machine learning model is generated by executing the federated learning model training using data from an identified second lower-level set of non-Byzantine client computers which are characterized as being non-Byzantine for the lower-level system. An intermediate global upper-level machine learning model is generated by executing the federated learning model training using data from an identified first upper-level set of non-Byzantine client computers which are characterized as being non-Byzantine for an upper-level system. A global upper-level machine learning model is generated by executing the federated learning model training using data from a second upper-level set of client computers which are characterized as being non-Byzantine for the upper-level system.

Systems and methods for feedback based instructional visual editing

Embodiments described herein provide a feedback based instructional image editing framework that employs a diffusion process to follow user instruction for image editing. A diffusion model is fine-tuned using a reward model, which may be trained via human annotation. The training of the reward model may be done by having the image editing model output a number of images, which a human annotator ranks based on their alignment with the original image and a given instruction.

Image guided surgical robotic platform

A system, including one or more processors configured to receive imaging data of a surgical site, where an end effector held by a surgical robot is positioned in the surgical site, determine one or more control parameters from the imaging data, and drive the surgical robot accordingly.

Distributed neural network training system

Apparatuses, systems, and techniques to train a machine-learned model. In at least one embodiment, a plurality of training clients each obtain an exclusive right to update a model in turn, and each client trains said model with training data not accessible to other training clients.

Contextual advertising through multimodal content analysis

A system and method for contextual advertising that analyzes video content through multimodal examination of visual, audio, and textual elements to create detailed contextual understanding of individual scenes.

Split keys for wallet recovery

Object recognition neural network training using multiple data sources

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an object recognition neural network using multiple data sources. One of the methods includes receiving training data that includes a plurality of training images from a first source and images from a second source. For each training image, contrast equalization is applied to generate a modified image, which is processed using the neural network to generate an object recognition output, and a loss is determined based on errors.

Systems and methods for multimodal pretraining for three-dimensional understanding models

A method of training a neural network based three-dimensional (3D) encoder is provided. A first plurality of samples of a training dataset are generated using a first 3D model. An image generator with multi-view rendering is used to generate a plurality of two-dimensional (2D) images having different viewpoints of the first 3D model. A first language model is used to generate a plurality of texts corresponding to the plurality of 2D images respectively. A first text for a first image is generated by using one or more text descriptions generated by the first language model. A point cloud is generated by randomly sampling points in the 3D model. The first plurality of samples are generated using the plurality of 2D images, the corresponding plurality of texts, and the point cloud. The neural network based 3D encoder is trained using the training dataset including the first plurality of samples.

Systems and methods for health analysis of a modeled software system

A method and apparatus are described for performing a health analysis of a modeled software system of a distributed services system. A stream of ledger transaction messages generated during execution of the software system are received and stored in a data store. A ledger data model is accessed, representing a state machine having states, data, and transitions between states indicative of proper functioning of the underlying software system. A ledger clearing operation generates monetary balances of states in the data model, and aggregated values associated with operational characteristics of various states.

Simulation modeling exchange

A simulation application container executes a simulation of a system in a simulation environment, through which an agent representing the system uses a reinforcement learning model to operate within the simulation environment.

System and method for force prediction

Apparatuses, systems, and techniques to predict forces associated with an object's surface. In at least one embodiment, forces associated with an object's surface are predicted using factorized implicit global convolution and one or more neural networks.

Cryptoasset custodial system with different rules governing access to logically separated cryptoassets

Methods, systems, and apparatus, including medium-encoded computer program products, for secure storage and retrieval of information, such as private keys, useable to control access to a blockchain, include, in at least one aspect, a method including: receiving a request to take an action with...

Apparatus, method and storage medium for data query

A question-answering system including a large language model stores history query information in a vector database and history query results in a scalar database; an incoming query is matched against the stored history so the result can be returned without re-running LLM inference, reducing latency and cost.

Context-sensitive linking of entities to private databases

Methods, systems, and computer-readable media for context-sensitive linking of entities to private databases are disclosed. An entity linking service stores a plurality of representations of entities.

Systems and methods for attention mechanism in three-dimensional object detection

Embodiments described herein provide a system for three-dimensional (3D) object detection. The system includes an input interface configured to obtain 3D point data describing spatial information of a plurality of points, and a memory storing a neural network based 3D object detection model having an encoder and a decoder. The system also includes processors to perform operations including: encoding, by the encoder, a first set of coordinates into a first set of point features and a set of object features; sampling a second set of point features from the first set of point features; generating, by attention layers at the decoder, a set of attention weights by applying cross-attention over at least the set of object features and the second set of point feature, and generate, by the decoder, a predicted bounding box among the plurality of points based on at least in part on the set of attention weights.

Systems and methods for open vocabulary instance segmentation in unannotated images

Embodiments described herein provide an open-vocabulary instance segmentation framework that adopts a pre-trained vision-language model to develop a pipeline in detecting novel categories of instances.

Real-time multi-user collaboration through authenticated data connections in notebook environment

A system enables real-time collaborative access to a notebook environment comprising a plurality of executable cells, including cells that retrieve data from a data warehouse. In response to a request from a first client device to access the notebook, the system checks permission and presents a first view; it then receives a request to initiate a collaboration session with a second user and sends an invitation to a second client device, keeping the users' views synchronized while preserving per-user data-connection credentials.

Automatic generation of an explanation, based on data lineage and user feedback, of a determination made by a machine learning system

A set of data for training a machine learning system can be modified to improve its performance. A message can be transmitted that includes an explanation of a determination made by the machine learning system.

Embedded predictive machine learning models

Data associated with one or more data sources is transformed into a format associated with a common ontology using one or more transformers. One or more machine learning models are generated based at least in part on the transformed data. The one or more machine learning models and the one or more transformers are provided to a remote device.

Secure reordering using tensor of indicators

An example system includes a processor to compute a tensor of indicators indicating a presence of partial sums in an encrypted vector of indicators. The processor can also securely reorder an encrypted array based on the computed tensor of indicators to generate a reordered encrypted array.

Feature Store with Integrated Tracking

The present application discloses a method, system, and computer system for managing a plurality of features and storing lineage information pertaining to the features. The method includes obtaining one or more datasets, determining a first feature, wherein the first feature is determined based at least in part on the one or more datasets, and storing the first feature in a feature store. The first feature is stored in association with a dataset indication of the one or more datasets from which the first feature is determined. The feature store comprises a plurality of features.

Feature Store with Integrated Tracking

The present application discloses a method, system, and computer system for managing a plurality of features and storing lineage information pertaining to the features. The method includes obtaining one or more datasets, determining a first feature, wherein the first feature is determined based at least in part on the one or more datasets, and storing the first feature in a feature store. The first feature is stored in association with a dataset indication of the one or more datasets from which the first feature is determined. The feature store comprises a plurality of features.

Elastic, ephemeral in-line deduplication service

A deduplication service can be provided to a storage domain from a services framework that expands and contracts to both meet service demand and to conform to resource management of a compute domain. The deduplication service maintains a fingerprint database and reference count data in compute domain resources, but persists these into the storage domain for use in the case of a failure or interruption of the deduplication service in the compute domain. The deduplication service responds to service requests from the storage domain with indications of paths in a user namespace and whether or not a piece of data had a fingerprint match in the fingerprint database. The indication of a match guides the storage domain to either store the piece of data into the storage backend or to reference another piece of data. The deduplication service uses the fingerprints to define paths for corresponding pieces of data.

Method, computer program product, and system for training a machine learning model to generate user embeddings and recipe embeddings in a common latent space for recommending one or more recipes to a user

Apparatus, method and storage medium for database pagination

An apparatus, method and storage medium for database pagination: acquire a pagination request, determine a query parameter, pagination pattern and request data amount, choose a target query mode, generate and run a query task, then page the target result set into multiple pages of pagination data.

Systems and Methods for Settlement Interfaces in Cellular Networks Using Blockchain Technology

The present system relates to settling transactions in a network environment using blockchain technology. The system comprises multiple network operator domains, each with network infrastructure components. A first operator domain includes a gateway for managing device connections and a blockchain interface for managing transaction settlements. A settlement management domain maintains transaction records for multiple network operators. The gateway manages device connections using WiFi, 4G, 5G, or 6G technologies, and the blockchain interface interacts with the blockchain for real-time settlement of transactions.

Systems and methods for resolving code vulnerabilities through collaborative agents

Systems and methods for resolving code vulnerabilities through collaborative agents which may include accessing a code base of an identified vulnerability; configuring a plurality of autonomous agents, each comprising a predefined agent role associated with application security remediation process; executing a directed workflow of the plurality of agents, wherein the workflow is a conditional sequence of agent-driven processing steps for generating a proposed resolution to the identified vulnerability; and outputting a candidate resolution for the vulnerability based on results produced by the workflow.

Page integrity assurance

Methods and systems are presented for authenticating web content. A hash value for each node of a first document object model (DOM) tree representing the web content is compared with the corresponding node of a second DOM tree associated with the host; the web content is authenticated when the trees match.

Systems and methods for biological metrics measurement

A wearable member may include a plurality of energy transmitters arranged on a surface of the wearable member and a plurality of energy receivers, multi-dimensionally arranged, each generating a signal from energy reflected by the tissue of the user.

Image processing neural networks with separable convolutional layers

Regenerating an interactive page based on current user interaction

In various embodiments, an optimization engine regenerates items included in an interactive page while the user is interacting with the interactive page. In operation, an optimization engine displays a portion of the interactive page during a viewing session.

On-the-fly calibration for improved on-device eye tracking

Calibration of eye tracking is improved by collecting additional calibration pairs while the user is using apps with eye tracking: a user input component is presented, a dwelling action is detected, and a calibration pair is recorded in response.

Fail-safe signal injection

A fail-safe signal injection system includes a fail-safe system controller that receives a control command and initiates to bypass a first system to inject a control signal into a second system responsive to the control command. The fail-safe signal injection system also includes a signal injection circuit implemented to inject the control signal into the second system responsive to power applied to the signal injection circuit, the signal injection circuit further implemented to fail-safe without the power applied, and thus operates to pass a control input from the first system through to the second system.