# US 2026/0019655 A1 · Machine learning model continuous training system

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0019655 A1
- Document ID: US20260019655A1
- Status: Published application
- Kind: A1
- Application number: 19/332,706
- Filed: 2025-09-18
- Published: 2026-01-15
- Earliest claimed: 2023-05-31
- Assignees: Snap Inc.
- Applicants: Snap Inc.
- Inventors: Kevin Sarabia Dela Rosa; Hao Hu; Yanjia Li
- Web page: https://patentgrep.com/patent/US20260019655A1

## CPC

- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- H04N 21/251: {Learning process for intelligent management, e.g. learning user preferences for recommending movies (details of learning user preferences for the retrieval of video data in a video database G06F16/739; computer systems using learning methods G06N3/08)}
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/0475: Generative networks
- G06N 3/048: Activation functions
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning
- G06N 3/09: Supervised learning
- G06N 3/091: Active learning
- G06N 3/092: Reinforcement learning
- G06N 3/096: Transfer learning
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
- G06V 20/41: {Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items (segmenting video sequences G06V20/49)}
- G06V 20/46: {Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames}
- H04N 21/23418: {involving operations for analysing video streams, e.g. detecting features or characteristics (television picture signal circuitry for scene change detection H04N5/147; filtering for image enhancement G06T5/00; methods or arrangements for recognising scenes G06V20/00; arrangements characterised by components specially adapted for monitoring, identification or recognition of video in broadcast systems H04H60/59)}
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

Described is a system for performing a set of machine learning model training operations that include: accessing media content items associated with interaction functions initiated by users of an interaction system, generating training data including labels for the media content items, extracting features from a media content item of the media content items, identifying additional media content items to include in the training data based on the extracted features from the media content item, processing the training data using a machine learning model to generate a media content item output; and updating one or more parameters of the machine learning model based on the media content item output. The system checks whether retraining criteria has been met, and repeats the set of machine learning model training operations to retrain the machine learning model.

## Claims

### Claim 1

A system comprising:
at least one processor;
at least one memory component storing instructions that, when executed by the at least one processor, cause the at least one processor to perform operations comprising:
accessing media content items associated with interaction functions initiated by users of an interaction system, wherein the media content items comprise content of the users posted on the interaction system enabling other users to view the posted media content items;
extracting features from at least one media content item of the media content items;
identifying additional media content items to include in the training data based on the extracted features from the at least one media content item;
generating labels for the additional media content items;
processing the training data and the labels using a machine learning model to generate a media content item output; and
updating one or more parameters of the machine learning model based on the media content item output to generate an updated machine learning model.

### Claim 2 (depends on 1)

The system of claim 1, wherein the content includes a plurality of images and a plurality of videos of the users posted on the interaction system enabling other users to view the posted media content items.

### Claim 3 (depends on 1)

The system of claim 1, wherein the content includes a plurality of content augmentations of the users posted on the interaction system enabling other users to view the posted media content items.

### Claim 4 (depends on 1)

The system of claim 1, wherein the operations further include, prior to extracting the features, generating the training data including labels for the media content items, wherein the labels are indicative of one or more characteristics of the media content items.

### Claim 5 (depends on 1)

The system of claim 1, wherein the operations further repeating the set of operations to retrain the updated machine learning model based on a retraining criterion being met.

### Claim 6 (depends on 1)

The system of claim 1, wherein the retraining criterion comprises a keyword indicative of a trend, wherein the operations further comprise tracking the use of one or more keywords in media content items and the retraining criterion includes meeting a threshold number of uses of the keyword.

### Claim 7 (depends on 1)

The system of claim 1, wherein identifying additional media content items comprises applying a distance metric to compare the media content item and individual additional media content items in order to identify the additional media content items.

### Claim 8 (depends on 7)

The system of claim 7, wherein extracting the features from the media content item comprises applying a machine learning model trained to extract features from one or more media content items, wherein generating the training data comprises adding the one or more extracted features to the labels.

### Claim 9 (depends on 7)

The system of claim 7, wherein the media content items comprise videos created by users to share with other users, wherein the features are extracted on a frame-by-frame basis.

### Claim 10 (depends on 7)

The system of claim 7, wherein the media content items comprise content augmentations created by users to share with other users, wherein the extracted features include the augmentations that are applied to a camera feed in real-time.

### Claim 11 (depends on 7)

The system of claim 7, wherein at least some of the media content items are in a different format than the additional media content items, wherein the media content items that are in the different format are compared with the additional media contents based on the extracted features.

### Claim 12 (depends on 1)

The system of claim 1, wherein the additional media content items were created by users in a different time period than when the accessed media content items were created.

### Claim 13 (depends on 1)

The system of claim 1, wherein the additional media content items were created by different users than the users that created the accessed media content items.

### Claim 14 (depends on 1)

The system of claim 1, wherein the additional media content items are identified based on metadata of the accessed media content items, wherein the metadata comprises a location where a user created the individual media content item.

### Claim 15 (depends on 1)

The system of claim 1, wherein the media content items comprise images or videos, and the interaction functions comprise media content items created by users and shared with other users.

### Claim 16 (depends on 15)

The system of claim 15, wherein the media content items that were created by users do not include labels for training the machine learning model, wherein generating the training data includes identifying keywords in captions of individual media content items or comments to the media content items from other users.

### Claim 17 (depends on 15)

The system of claim 15, wherein the media content items comprise content augmentations that add interactive digital elements in real-time to a camera feed.

### Claim 18 (depends on 1)

The system of claim 1, wherein the operations further comprise adding the training data to an existing set of training data, wherein repeating the set of operations further comprises adding newly accessed training data to the existing set of training data such that the existing set of training data increases in size with each repeating of the set of operations, wherein processing the training data using the machine learning model to generate the media content item output further comprises processing the existing set of training data using the machine learning model.

### Claim 19

A method comprising:
accessing media content items associated with interaction functions initiated by users of an interaction system, wherein the media content items comprise content of the users posted on the interaction system enabling other users to view the posted media content items;
extracting features from at least one media content item of the media content items;
identifying additional media content items to include in training data based on the extracted features from the at least one media content item;
generating labels for the additional media content items;
processing the training data and the labels using a machine learning model to generate a media content item output; and
updating one or more parameters of the machine learning model based on the media content item output to generate an updated machine learning model.

### Claim 20

A non-transitory computer-readable storage medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform operations comprising:
accessing media content items associated with interaction functions initiated by users of an interaction system, wherein the media content items comprise content of the users posted on the interaction system enabling other users to view the posted media content items;
extracting features from at least one media content item of the media content items;
identifying additional media content items to include in training data based on the extracted features from the at least one media content item;
generating labels for the additional media content items;
processing the training data and the labels using a machine learning model to generate a media content item output; and
updating one or more parameters of the machine learning model based on the media content item output to generate an updated machine learning model.
