# US 2026/0065072 A1 · Knowledge acquisition for distributed learning models

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

- Number: US 2026/0065072 A1
- Document ID: US20260065072A1
- Status: Published application
- Kind: A1
- Application number: 18/963,976
- Filed: 2024-11-29
- Published: 2026-03-05
- Earliest claimed: 2024-09-05
- Assignees: Dell Products L.P.
- Applicants: Dell Products LP.
- Inventors: William Jeffery White; Xin Ma; George Currie
- Web page: https://patentgrep.com/patent/US20260065072A1

## CPC

- G06F 16/285: {Clustering or classification}
- G06N 3/045: Combinations of networks
- G06N 3/098: Distributed learning, e.g. federated learning
- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06N 5/04: Inference or reasoning models
- G06N 5/043: Distributed expert systems; Blackboards
- G06Q 10/087: Inventory or stock management, e.g. order filling, procurement or balancing against orders
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

The disclosure includes a distributed learning knowledge acquirer. The distributed learning knowledge acquirer allows for the training of discriminative AI models on the output data from generative AI models in a local small foundation system and a centralized large foundation system. This increases the effectiveness of the discriminative AI models which are used in an artificial intelligence hybrid distributed system to make first classifications on input data. Classifying the input data completely at the discriminative AI model level is cheaper, easier, and more time efficient making updates of the discriminative AI model critical. The distributed learning knowledge acquirer is connected to a centralized system allowing for data sharing between distributed learning knowledge acquirers in different edge sites up to and including federated learning.

## Claims

### Claim 1

A method for creating discriminative artificial intelligence (AI) models, comprising:
receiving, at a distributed learning knowledge acquirer, data from a local small foundation system, a distributed endpoint management system, and a centralized large foundation system, wherein:
the distributed learning knowledge acquirer is connected to the distributed endpoint management system and the local small foundation system which are located on a distributed edge site of a plurality of distributed edge sites,
the plurality of distributed edge sites are connected to a centralized system to form an artificial intelligence (AI) hybrid distributed inference system, and
the centralized large foundation system is located on the centralized system with an external information repository;
making a determination that the data to determine reaches a threshold to perform a drift evaluation;
based on the determination, evaluating the data to determine the data causes a conceptual drift and can be used as local training data;
transmitting the local training data to a centralized training dataset repository of the external information repository and a local federated model trainer;
training a local discriminative model using the local training data;
evaluating the local discriminative model using a model evaluation module to determine local discriminative model is ready to be used at the distributed endpoint management system; and
transmitting the local discriminative model to a discriminative AI model repository of the external information repository and the distributed endpoint management system, wherein the local discriminative model includes elements trained on data produced locally on the distributed edge site by classifications from the local small foundation system and data produced centrally at the centralized system by classifications of the centralized large foundation system.

### Claim 2 (depends on 1)

The method of claim 1, wherein each distributed edge site is in a factory of a plurality of factories.

### Claim 3 (depends on 2)

The method of claim 2, wherein the local discriminative model is used at the distributed endpoint management system to perform classifications on input data from the factory.

### Claim 4 (depends on 1)

The method of claim 1, wherein the data produced locally on the distributed edge site by classifications from the local small foundation system causes the discriminative AI models to be customized in the distributed edge site or other distributed site from other distributed edge sites.

### Claim 5 (depends on 1)

The method of claim 1, wherein the local small foundation system has less than 10 billion parameters.

### Claim 6 (depends on 1)

The method of claim 1, further comprising:
prior to training the local discriminative model:
transmitting the data to a digital twin model to rank the data for use in training the local discriminative model; and
receiving a ranking of the data from the digital twin model.

### Claim 7 (depends on 1)

The method of claim 1, wherein the local training data is distributed by the centralized training dataset repository to a second distributed learning knowledge acquirer in a second distributed edge site of the plurality of distributed edge sites to train and evaluate discriminative AI models.

### Claim 8 (depends on 1)

The method of claim 1, wherein evaluating the local discriminative model comprises using external training data from a second distributed edge site transmitted to the distributed learning knowledge acquirer from the centralized training dataset repository.

### Claim 9

A non-transitory computer readable medium comprising computer readable program code, which when executed by a computer processor enables the computer processor to perform a method for creating discriminative artificial intelligence (AI) models, the method comprising:
receiving, at a distributed learning knowledge acquirer, data from a local small foundation system, a distributed endpoint management system, and a centralized large foundation system, wherein:
the distributed learning knowledge acquirer is connected to the distributed endpoint management system and the local small foundation system which are located on a distributed edge site of a plurality of distributed edge sites,
the plurality of distributed edge sites are connected to a centralized system to form an artificial intelligence (AI) hybrid distributed inference system, and
the centralized large foundation system is located on the centralized system with an external information repository;
making a determination that the data to determine reaches a threshold to perform a drift evaluation;
based on the determination, evaluating the data to determine the data causes a conceptual drift and can be used as local training data;
transmitting the local training data to a centralized training dataset repository of the external information repository and a local federated model trainer;
training a local discriminative model using the local training data;
evaluating the local discriminative model using a model evaluation module to determine local discriminative model is ready to be used at the distributed endpoint management system; and
transmitting the local discriminative model to a discriminative AI model repository of the external information repository and the distributed endpoint management system, wherein the local discriminative model includes elements trained on data produced locally on the distributed edge site by classifications from the local small foundation system and data produced centrally at the centralized system by classifications of the centralized large foundation system.

### Claim 10 (depends on 9)

The non-transitory computer readable medium of claim 9, wherein each distributed edge site is in a factory of a plurality of factories.

### Claim 11 (depends on 10)

The non-transitory computer readable medium of claim 10, wherein the local discriminative model is used at the distributed endpoint management system to perform classifications on input data from the factory.

### Claim 12 (depends on 9)

The non-transitory computer readable medium of claim 9, wherein the data produced locally on the distributed edge site by classifications from the local small foundation system causes the discriminative AI models to be customized in the distributed edge site or other distributed site from other distributed edge sites.

### Claim 13 (depends on 9)

The non-transitory computer readable medium of claim 9, wherein the local small foundation system has less than 10 billion parameters.

### Claim 14 (depends on 9)

The non-transitory computer readable medium of claim 9, wherein the local training data is distributed by the centralized training dataset repository to a second distributed learning knowledge acquirer in a second distributed edge site of the plurality of distributed edge sites to train and evaluate discriminative AI models.

### Claim 15 (depends on 9)

The non-transitory computer readable medium of claim 9, wherein evaluating the local discriminative model comprises using external training data from a second distributed edge site transmitted to the distributed learning knowledge acquirer from the centralized training dataset repository.

### Claim 16

A distributed learning knowledge acquirer, comprising:
a processor;
storage comprising instructions, which when executed by the processor perform a method, the method comprising:
receiving, at the distributed learning knowledge acquirer, data from a local small foundation system, a distributed endpoint management system, and a centralized large foundation system, wherein:
the distributed learning knowledge acquirer is connected to the distributed endpoint management system and the local small foundation system which are located on a distributed edge site of a plurality of distributed edge sites,
the plurality of distributed edge sites are connected to a centralized system to form an artificial intelligence (AI) hybrid distributed inference system, and
the centralized large foundation system is located on the centralized system with an external information repository;
making a determination that the data to determine reaches a threshold to perform a drift evaluation;
based on the determination, evaluating the data to determine the data causes a conceptual drift and can be used as local training data;
transmitting the local training data to a centralized training dataset repository of the external information repository and a local federated model trainer;
training a local discriminative model using the local training data;
evaluating the local discriminative model using a model evaluation module to determine local discriminative model is ready to be used at the distributed endpoint management system; and
transmitting the local discriminative model to a discriminative AI model repository of the external information repository and the distributed endpoint management system, wherein the local discriminative model includes elements trained on data produced locally on the distributed edge site by classifications from the local small foundation system and data produced centrally at the centralized system by classifications of the centralized large foundation system.

### Claim 17 (depends on 16)

The distributed learning knowledge acquirer of claim 16, wherein each distributed edge site is in a factory of a plurality of factories.

### Claim 18 (depends on 17)

The distributed learning knowledge acquirer of claim 17, wherein the local discriminative model is used at the distributed endpoint management system to perform classifications on input data from the factory.

### Claim 19 (depends on 16)

The distributed learning knowledge acquirer of claim 16, wherein the local training data is distributed by the centralized training dataset repository to a second distributed learning knowledge acquirer in a second distributed edge site of the plurality of distributed edge sites to train and evaluate discriminative AI models.

### Claim 20 (depends on 16)

The distributed learning knowledge acquirer of claim 16, wherein evaluating the local discriminative model comprises using external training data from a second distributed edge site transmitted to the distributed learning knowledge acquirer from the centralized training dataset repository.
