Topological Sparse Training Process for Machine Learning Models
Summary
USPTO published patent application US20260099725A1 for a topological sparse training process for machine learning models. The application covers methods for executing ML models with attention heads, using loss functions configured for preferential attachment of neurons, and modifying parameters including setting values to zero to generate sparse ML models. The invention aims to optimize neural network architecture through preferential attachment mechanisms during training.
What changed
USPTO published patent application US20260099725A1 for a topological sparse training process for machine learning models. The application covers methods including executing ML models with attention heads on training data, determining differences using loss functions configured for preferential attachment of neurons, modifying parameter values including setting parameters to zero to generate sparse ML models, and executing the sparse model on additional training data.
Affected parties including ML researchers, AI developers, and technology companies may reference this patent application to understand emerging techniques in neural network optimization and sparse model training. The patent does not impose compliance obligations but provides public notice of the claimed invention and its technical approach.
Archived snapshot
Apr 17, 2026GovPing captured this document from the original source. If the source has since changed or been removed, this is the text as it existed at that time.
TOPOLOGICAL SPARSE TRAINING PROCESS FOR MACHINE LEARNING MODELS
Application US20260099725A1 Kind: A1 Apr 09, 2026
Inventors
Amit Dhurandhar, Soham Dan, Aurelie Chloe Lozano, Georgios Kollias, Ronny Luss, Payel Das, Tejaswini Pedapati
Abstract
An example operation may include one or more of executing a machine learning (ML) model with a plurality of attention heads on a training data input during an epoch to generate a predicted output, determining a difference between the predicted output an and actual output corresponding to the training data input based on a loss function that is configured to perform preferential attachment of neurons in the ML model, modifying parameter values of the ML model based on the difference, wherein the modifying comprises modifying at least one parameter value of the parameter values of the ML model to be set to zero to generate a sparse ML model, and executing the sparse ML model on an additional training data input during an additional epoch to generate an additional predicted output.
CPC Classifications
G06N 3/096 G06N 3/0475
Filing Date
2024-10-03
Application No.
18906079
Related changes
Get daily alerts for USPTO Patent Applications - AI & Computing (G06N)
Daily digest delivered to your inbox.
Free. Unsubscribe anytime.
Source
About this page
Every important government, regulator, and court update from around the world. One place. Real-time. Free. Our mission
Source document text, dates, docket IDs, and authority are extracted directly from USPTO.
The summary, classification, recommended actions, deadlines, and penalty information are AI-generated from the original text and may contain errors. Always verify against the source document.
Classification
Who this affects
Taxonomy
Browse Categories
Get alerts for this source
We'll email you when USPTO Patent Applications - AI & Computing (G06N) publishes new changes.
Subscribed!
Optional. Filters your digest to exactly the updates that matter to you.