IP Library Granted Patent US 12,689,389
Granted Patent B2
US 12,689,389 · App. 18/440,948 · Granted Jul 21, 2026

Systems and methods for compression of artificial intelligence

Inventors: Jonghyeon Kim (San Jose, CA); Soogil Jeong (Pleasanton, CA)
Assignee: SAMSUNG ELECTRONICS CO., LTD.
H03M7/702
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,689,389
App. No.
18/440,948
Granted
Jul 21, 2026
Kind
B2
Abstract

Provided are systems, methods, and apparatuses for compression of artificial intelligence models. In one or more examples, the systems, devices, and methods include categorizing data based on an analysis of a distribution of the data, generating compressed data based on the data and on a compression algorithm that is selected based on the categorization, and storing the compressed data in a storage device. In one or more examples, the systems, devices, and methods include identifying an address associated with compressed data based on a request for the compressed data, determining a decompression algorithm based on the address, and decompressing the compressed data using the determined decompression algorithm.

Claims (36)

1 . A method comprising:

categorizing data based on an analysis of a distribution of the data in a histogram, wherein the data includes weights of an artificial intelligence model;

generating compressed data based on the data and on a compression algorithm that is selected based on the categorization; and

storing the compressed data in a storage device.

2 . The method of claim 1 , further comprising generating address boundary information based on the compression algorithm that is selected.

3 . The method of claim 2 , further comprising storing a mapping between the address boundary information and the compression algorithm.

4 . The method of claim 2 , wherein the address boundary information is based on an address where the compressed data is stored in the storage device.

5 . The method of claim 1 , further comprising generating a histogram based on the data.

6 . The method of claim 1 , wherein the data includes weights of an artificial intelligence model, and further comprising generating a histogram based on the weights of the artificial intelligence model.

7 . The method of claim 6 , wherein the distribution of the data is based on the histogram.

8 . The method of claim 1 , wherein:

the storage device includes a solid-state drive, and

at least a portion of the compression is performed by a processing unit communicatively connected to the storage device.

9 . A method comprising:

identifying an address associated with compressed data based on a request for the compressed data;

determining a decompression algorithm based on the address, wherein the address is associated with an address range of where the compressed data is stored in a storage device; and

decompressing the compressed data using the determined decompression algorithm.

10 . The method of claim 9 , wherein identifying the decompression algorithm is based on comparing the address of the compressed data to a boundary address associated with the decompression algorithm and identifying a match based on the comparing.

11 . The method of claim 9 , wherein the compressed data includes weights of an artificial intelligence model.

12 . The method of claim 9 , wherein:

the storage device includes a solid-state drive, and

at least a portion of the compression is performed by a processing unit communicatively connected to the storage device.

13 . The method of claim 9 , wherein the storage device includes a mapping between a boundary address and the decompression algorithm.

14 . A device, comprising:

at least one processing unit to:

identify an address associated with compressed data based on a request for the compressed data; and

determine a decompression algorithm based on the address, wherein the address is associated with an address range of where the compressed data is stored in a storage device; and

a decompressor to decompress the compressed data using the determined decompression algorithm.

15 . The device of claim 14 , wherein:

the decompression algorithm corresponds to a distribution of data that includes weights of an artificial intelligence model, and

the compressed data includes the weights of the artificial intelligence model.

16 . The device of claim 14 , wherein, to identify the decompression algorithm, the at least one processing unit is configured to:

compare the address of the compressed data to a boundary address associated with the decompression algorithm; and

identify a match based on the comparison.

17 . The device of claim 14 , wherein the storage device stores a mapping between a boundary address and the decompression algorithm.

18 . The device of claim 14 , wherein the storage device includes a solid-state drive.