# US 2025/0095217 A1 · Low rank matrix compression

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

- Number: US 2025/0095217 A1
- Document ID: US20250095217A1
- Status: Published application
- Kind: A1
- Application number: 18/903,291
- Filed: 2024-10-01
- Published: 2025-03-20
- Earliest claimed: 2017-04-08
- Assignees: Intel Corporation
- Applicants: Intel Corporation
- Inventors: Tomer Bar-On; Jacob Subag; Yaniv Fais; Jeremie Dreyfuss; Gal Novik; Gal Leibovich; Tomer Schwartz; Ehud Cohen; Lev Faivishevsky; Uzi Sarel; Amitai Armon; Yahav Shadmiy
- Web page: https://patentgrep.com/patent/US20250095217A1

## CPC

- G06T 9/002: {using neural 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 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/048: Activation functions
- G06N 3/0495: Quantised networks; Sparse networks; Compressed networks
- 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 3/098: Distributed learning, e.g. federated learning
- H04N 19/42: characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation (H04N19/635 takes precedence)
- H04N 19/436: using parallelised computational arrangements
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION

## Abstract

In an example, an apparatus comprises logic, at least partially including hardware logic, to implement a lossy compression algorithm which utilizes a data transform and quantization process to compress data in a convolutional neural network (CNN) layer. Other embodiments are also disclosed and claimed.

## Claims

### Claim 1

An apparatus comprising:
processing circuitry to:
facilitate lossy compression algorithm, based on data transform and quantization process, to compress data in a convolutional neural network (CNN) layer.

### Claim 2 (depends on 1)

The apparatus of claim 1, wherein the processing circuitry is further to compress one or more weights in a convolutional neural network (CNN) layer in a frequency domain, wherein the one or more weights are quantized in the frequency domain.

### Claim 3

(canceled)

### Claim 4 (depends on 2)

The apparatus of claim 2, wherein the processing circuitry is further to:
decompress the data in the convolutional neural network (CNN) layer.

### Claim 5 (depends on 2)

The apparatus of claim 2, wherein the processing circuitry is further to:
apply an inversed transform to the convolutional neural network (CNN) layer before saving the compute in one or more processing resources, wherein the processing circuitry is coupled to a memory, the processing circuitry comprising one or more of graphics processing circuitry or application processing circuitry.

### Claim 6

(canceled)

### Claim 7

(canceled)

### Claim 8

(canceled)

### Claim 9

(canceled)

### Claim 10

(canceled)

### Claim 11

A method comprising:
facilitating, by a processor of a computing device, lossy compression, based on a data transform and quantization process, to compress data in a convolutional neural network (CNN) layer.

### Claim 12 (depends on 11)

The method of claim 11, further comprising compressing one or more weights in a convolutional neural network (CNN) layer in a frequency domain, wherein the one or more weights are quantized in the frequency domain.

### Claim 13 (depends on 12)

The method of claim 12, further comprising decompressing the data in the convolutional neural network (CNN) layer.

### Claim 14 (depends on 12)

The method of claim 12, further comprising applying an inversed transform to the convolutional neural network (CNN) layer before saving the compute in one or more processing resources, wherein the processor is coupled to a memory, the processor comprising one or more of a graphics processor or an application processor.

### Claim 15

At least one computer-readable medium having stored thereon instructions which, when executed, cause a computing device to perform operations comprising:
facilitating lossy compression, based on a data transform and quantization process, to compress data in a convolutional neural network (CNN) layer.

### Claim 16 (depends on 15)

The computer-readable medium of claim 15, wherein the operations further comprise compressing one or more weights in a convolutional neural network (CNN) layer in a frequency domain, wherein the one or more weights are quantized in the frequency domain.

### Claim 17 (depends on 16)

The computer-readable medium of claim 16, wherein the operations further comprise decompressing the data in the convolutional neural network (CNN) layer.

### Claim 18 (depends on 16)

The computer-readable medium of claim 16, wherein the operations further comprise applying an inversed transform to the convolutional neural network (CNN) layer before saving the compute in one or more processing resources, wherein the computing device comprises one or more processors comprising one or more graphics processors or one or more application processors.
