# US 2020/0351386 A1 · Methods and Apparatus for Compression Data Streams

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

- Number: US 2020/0351386 A1
- Document ID: US20200351386A1
- Status: Published application
- Kind: A1
- Application number: 16/842,478
- Filed: 2020-04-07
- Published: 2020-11-05
- Earliest claimed: 2019-04-30
- Assignees: Marvell Asia Pte Ltd
- Applicants: Marvell Asia Pte, Ltd.
- Inventors: KALPENDU RATANSHI PASAD; HONG JIK KIM
- Web page: https://patentgrep.com/patent/US20200351386A1

## CPC

- H03M 7/40: Conversion to or from variable length codes, e.g. Shannon-Fano code, Huffman code, Morse code
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- H03M 7/30: Compression (speech analysis-synthesis for redundancy reduction G10L19/00; for image communication H04N); Expansion; Suppression of unnecessary data, e.g. redundancy reduction
- H04L 69/04: Protocols for data compression, e.g. ROHC
- G06N 3/045: Combinations of networks
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- H03M 7/3082: {Vector coding (for television signals, see H04N19/94)}
- H03M: CODING; DECODING; CODE CONVERSION IN GENERAL (using fluidic means F15C4/00; optical analogue/digital converters G02F7/00; coding, decoding or code conversion, specially adapted for particular applications, see the relevant subclasses, e.g. G01D, G01R, G06F, G06T, G09G, G10L, G11B, G11C, H04B, H04L, H04M, H04N; ciphering or deciphering for cryptography or other purposes involving the need for secrecy G09C)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

Methods and apparatus for compressing data streams. In an embodiment, a method includes calculating a probability distribution function (PDF) for scaler data, matching the PDF to PDF templates to determine a closest matching PDF template, and selecting an encoder corresponding to the closest matching PDF template wherein a corresponding encoder identifier is determined. The method also includes encoding the scaler data with the encoder to generate an encoded stream, and transmitting the encoded stream and the encoder identifier.

## Claims

### Claim 1

A method for encoding information, the method comprising:
calculating a probability distribution function (PDF) for scaler data;
matching the PDF to PDF templates to determine a closest matching PDF template;
determining an encoder corresponding to the closest matching PDF template, wherein a corresponding encoder identifier is determined;
encoding the scaler data with the encoder to generate an encoded stream; and
transmitting the encoded stream and the encoder identifier.

### Claim 2 (depends on 1)

The method of claim 1, wherein the operation of determining the encoder comprises determining an entropic encoder corresponding to the closest matching PDF template.

### Claim 3 (depends on 1)

The method of claim 1, further comprising maintaining the PDF templates in a database.

### Claim 4 (depends on 3)

The method of claim 3, wherein the operation of matching comprises:
accessing the database to obtain the PDF templates; and
matching the PDF to the PDF templates using KL (Kullback-Leibler) divergence to determine the closest matching PDF template.

### Claim 5 (depends on 1)

The method of claim 1, wherein the operation of transmitting further comprising transmitting the encoded stream and the encoder identifier over at least one of a wired link and a wireless link.

### Claim 6 (depends on 1)

The method of claim 1, further comprising:
receiving the encoded stream and the encoder identifier;
determining a decoder based on the encoder identifier;
decoding the encoded stream using the decoder to obtain the scaler data.

### Claim 7

A method for encoding information, the method comprising:
binning received vector data into a plurality of bins;
determining a centroid for each bin to form a dictionary of centroids;
tagging the vector data with tags that identify a centroid associated with each vector;
counting a number of vectors in each bin to construct a probability distribution function (PDF) of centroids;
matching the PDF of centroids to PDF templates to determine a closest matching PDF template;
determining an encoder corresponding to the closest matching PDF template wherein a corresponding encoder identifier is identified;
encoding the tags with the encoder to generate an encoded data stream; and
transmitting the encoded stream, the encoder identifier, and the dictionary of centroids.

### Claim 8 (depends on 7)

The method of claim 7, wherein the operation of encoding comprises encoding the tags with an entropic encoder to generate the encoded data stream,

### Claim 9 (depends on 7)

The method of claim 7, further comprising optionally encoding the vector data with an auto-encoder prior to the operation of determining the vector centroids.

### Claim 10 (depends on 7)

The method of claim 7, further comprising maintaining the PDF templates in a database.

### Claim 11 (depends on 7)

The method of claim 7, wherein the operation of matching comprises:
accessing the database to obtain the PDF templates; and
matching the PDF of centroids to the PDF templates using KL (Kullback-Leibler) divergence to determine the closest matching PDF template.

### Claim 12 (depends on 7)

The method of claim 7, wherein the operation of transmitting further comprising transmitting the encoded stream and the encoder identifier over at least one of a wired link and a wireless link.

### Claim 13 (depends on 7)

The method of claim 7, further comprising:
receiving the encoded stream, the encoder identifier, and the dictionary;
determining a decoder based on the encoder identifier;
decoding the encoded stream using the decoder to obtain the tags.
converting the tags into a vector representation using the dictionary; and
outputting the vector representation as the vector data.

### Claim 14 (depends on 13)

The method of claim 13, further comprising optionally decoding the vector representation using a decoder portion of an auto-encoder before the operation of outputting.

### Claim 15

An apparatus for encoding information, the apparatus comprising:
a probability distribution function (PDF) computation circuit that calculates a PDF associated with input data;
a matching circuit that matches the PDF to PDF templates to determine a closest matching PDF template, and determines an encoder corresponding to the closest matching PDF template wherein a corresponding encoder identifier is determined; and
an encoding circuit that encodes input data with the encoder to generate an encoded stream, and transmits the encoded stream and the encoder identifier.

### Claim 16 (depends on 15)

The apparatus of claim 15, further comprising a database that stores the PDF templates.

### Claim 17 (depends on 15)

The apparatus of claim 15, wherein the matching circuit accesses the database to obtain the PDF templates, and matches the PDF to the PDF templates using KL (Kullback-Leibler) divergence to determine the closest matching PDF template.

### Claim 18 (depends on 15)

The apparatus of claim 15, wherein the encoding circuit transmits the encoded stream and the encoder identifier over at least one of a wired link and a wireless link.

### Claim 19 (depends on 15)

The apparatus of claim 15, wherein the input data is scaler data and the apparatus further comprises:
a receiver that receives the encoded stream and the encoder identifier; and
a decoding circuit that determines a decoder based on the encoder identifier, and decodes the encoded stream using the decoder to obtain the scaler data.

### Claim 20 (depends on 15)

The apparatus of claim 15, wherein the input data is vector data and the apparatus further comprises a vector centroid computation circuit that generates tagged vectors from the vector data, and wherein the tagged vectors are input to the probability distribution function (PDF) computation circuit as the input data.
