# US 2026/0228505 A1 · Ai based synthetic compute workload generation and characterization

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

- Number: US 2026/0228505 A1
- Document ID: US20260228505A1
- Status: Published application
- Kind: A1
- Application number: 19/178,577
- Filed: 2025-04-14
- Published: 2026-08-06
- Earliest claimed: 2025-02-05
- Assignees: Hewlett Packard Enterprise Development LP
- Applicants: Hewlett Packard Enterprise Development LP
- Inventors: Susanta Adhikary; Ravindhar Uppada
- Web page: https://patentgrep.com/patent/US20260228505A1

## CPC

- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/0475: Generative networks
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/063: using electronic means
- G06N 3/08: Learning methods
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- G06N 3/09: Supervised learning
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 20/00: Machine learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Systems and methods are provided for generating synthetic workload data based on real workload data using a first machine learning model. In some examples, the system may classify, using a second machine learning model, the synthetic workload data. The classification may help identify and adjust a system configuration setting of a compute device and, in some examples, retrain the second machine learning model.

## Claims

### Claim 1

A computer-implemented method comprising:
receiving first real workload data of a compute device;
generating, using a first machine learning model, synthetic workload data associated with the first real workload data that includes random noise or conditional input;
in response to generating a threshold amount of the synthetic workload data as output from the first machine learning model, classifying, using a second machine learning model, the synthetic workload data as a workload type;
identifying and adjusting a system configuration setting of the compute device based on the workload type; and
in response to receiving second real workload data that is adjusted based on the system configuration setting, retraining the second machine learning model with the second real workload data.

### Claim 2 (depends on 1)

The method of claim 1, wherein the first machine learning model is a conditional Generative Adversarial Network (cGAN).

### Claim 3 (depends on 1)

The method of claim 1, further comprising:
in response to the classification of the synthetic workload data, enabling an assignment of a Basic Input/Output System (BIOS) workload profile for the compute device, wherein the BIOS workload profile optimizes system configurations for specific CPU-bound, memory-bound, and I/O-bound tasks.

### Claim 4 (depends on 1)

The method of claim 1, wherein the first real workload data of the compute device comprises CPU, memory, and I/O utilization, and the workload type comprises CPU-bound, memory-bound, and I/O-bound tasks.

### Claim 5 (depends on 1)

The method of claim 1, wherein the first machine learning model is selected based on the workload type, and the first machine learning model is trained based on the first real workload data.

### Claim 6 (depends on 1)

The method of claim 1, wherein the second machine learning model is a Long Short-Term Memory (LSTM) model.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
quantifying the synthetic workload data generated by the first machine learning model using Mean Squared Error (MSE) with a formula comprising:
MSE
=
1
n
⁢
∑
i
=
1
n
(
y
i
-
y
^
i
)
2
where n is a total number of data points, yi is the first real workload data for the i-th sample, and ŷi is the synthetic workload data for the i-th sample.

### Claim 8 (depends on 1)

The method of claim 1, wherein the workload type is CPU bound, memory bound, or I/O bound tasks.

### Claim 9

A compute component comprising:
a memory storing instructions; and
a processor communicatively coupled to the memory and configured to execute the instructions to:
receive first real workload data of a compute device associated with processor tasks of the compute device;
generate, using a first machine learning model, synthetic workload data associated with the first real workload data that includes random noise or conditional input;
in response to generating a threshold amount of the synthetic workload data as output from the first machine learning model, classify, using a second machine learning model, the synthetic workload data as being associated with the processor tasks of the compute device;
identify and adjust a system configuration setting of a processor of the compute device; and
in response to receiving second real workload data that is adjusted based on the system configuration setting, retrain the second machine learning model with the second real workload data.

### Claim 10 (depends on 9)

The compute component of claim 9, wherein the first machine learning model is a conditional Generative Adversarial Network (cGAN).

### Claim 11 (depends on 9)

The compute component of claim 9, wherein the processor is further configured to execute the instructions to:
enable an assignment of a Basic Input/Output System (BIOS) workload profile for the compute device, wherein the BIOS workload profile optimizes system configurations for specific CPU-bound, memory-bound, and I/O-bound tasks.

### Claim 12 (depends on 9)

The compute component of claim 9, wherein the first real workload data of the compute device comprises CPU, memory, and I/O utilization, and the workload type comprises CPU-bound, memory-bound, and I/O-bound tasks.

### Claim 13 (depends on 9)

The compute component of claim 9, wherein the first machine learning model is selected based on the workload type, and the first machine learning model is trained based on the first real workload data.

### Claim 14 (depends on 9)

The compute component of claim 9, wherein the second machine learning model is a Long Short-Term Memory (LSTM) model.

### Claim 15 (depends on 9)

The compute component of claim 9, wherein the processor is further configured to execute the instructions to:
quantify the synthetic workload data generated by the first machine learning model using Mean Squared Error (MSE) with a formula comprising:
MSE
=
1
n
⁢
∑
i
=
1
n
(
y
i
-
y
^
i
)
2
where n is a tc number of data points, yi is the first real workload data for the i-th sample, and is the synthetic workload data for the i-th sample.

### Claim 16 (depends on 9)

The compute component of claim 9, wherein the workload type is CPU bound, memory bound, or I/O bound tasks.

### Claim 17

A server comprising:
a memory storing instructions; and
a processor communicatively coupled to the memory and configured to execute the instructions to:
receive first real workload data of a compute device;
generate, using a conditional Generative Adversarial Network (cGAN) model, synthetic workload data associated with the first real workload data that includes random noise or conditional input;
in response to generating a threshold amount of the synthetic workload data as output from the cGAN model, classify, using a Long Short-Term Memory (LSTM) model, the synthetic workload data as a workload type;
identify and adjust a system configuration setting of the compute device based on the workload type; and
in response to receiving second real workload data that is adjusted based on the system configuration setting, retrain the LSTM model with the second real workload data.

### Claim 18 (depends on 17)

The server of claim 17, wherein the processor is further configured to execute the instructions to:
enable an assignment of a Basic Input/Output System (BIOS) workload profile for the compute device, wherein the BIOS workload profile optimizes system configurations for specific CPU-bound, memory-bound, and I/O-bound tasks.

### Claim 19 (depends on 17)

The server of claim 17, wherein the first real workload data of the compute device comprises CPU, memory, and I/O utilization, and the workload type comprises CPU-bound, memory-bound, and I/O-bound tasks.

### Claim 20 (depends on 17)

The server of claim 17, wherein the processor is further configured to execute the instructions to:
quantify the synthetic workload data generated by the first machine learning model using Mean Squared Error (MSE) with a formula comprising:
MSE
=
1
n
⁢
∑
i
=
1
n
(
y
i
-
y
^
i
)
2
where n is a total number of data points, yi is the first real workload data for the i-th sample, and yi is the synthetic workload data for the i-th sample.
