# US 2021/0344530 A1 · Receiver and transmitter adaptation using stochastic gradient hill climbing with genetic mutation

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

- Number: US 2021/0344530 A1
- Document ID: US20210344530A1
- Status: Published application
- Kind: A1
- Application number: 17/326,032
- Filed: 2021-05-20
- Published: 2021-11-04
- Earliest claimed: 2019-05-22
- Assignees: NVIDIA Corporation
- Applicants: Nvidia Corporation
- Inventors: Vishnu Balan; Mohammad Mobin; Rohit RATHI; Dai DAI
- Web page: https://patentgrep.com/patent/US20210344530A1

## CPC

- H04L 25/03146: {with a recursive structure (H04L25/03127 takes precedence)}
- H04L 27/01: Equalisers {(baseband equalizers at the transmitter end H04L25/03343; in analogue transmission systems H04B3/04, H04B7/005)}
- H04L 25/03878: {Line equalisers; line build-out devices}
- H04L 2025/03636: {Algorithms using least mean square [LMS]}
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

A receiver receives communications over a communication channel, which may distort an incoming communication signal. In order to counter this distortion, the frequency response of the receiver is manipulated by adjusting several frequency response parameters. Each frequency response parameter controls at least a portion of the frequency response of the receiver. The optimal values for the frequency response parameters are determined by modifying an initial set of values for the frequency response parameters through one or more of stochastic hill climbing operations until a performance metric associated with the receiver reaches a local maximum. The modified values are displaced through one or more mutation operations. The stochastic hill climbing operations may subsequently be performed on the mutated values to generate the final values for the frequency response parameters.

## Claims

### Claim 1

A computer-implemented method for equalizing a transmitter, the method comprising:
performing one or more stochastic gradient hill climbing operations and one or more genetic mutation operations on one or more parameters used to equalize a signal transmitted by the transmitter;
determining that, in response to the one or more stochastic gradient hill climbing operations and the one or more genetic mutation operations, a figure of merit of an eye diagram associated with a frequency response to the signal at a receiver has reached a local maximum;
determining, when the figure of merit is at the local maximum, one or more values for the one or more parameters; and
transmitting, to the transmitter, the one or more values for the one or more parameters, wherein at least one equalization operation is performed on the signal at the transmitter based on the one or more values for the one or more parameters.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, wherein the one or more values for the one or more parameters are transmitted via a first channel that is separate and distinct from a second channel via which the signal is transmitted by the transmitter.

### Claim 3 (depends on 1)

The computer-implemented method of claim 1, wherein the one or more parameters are associated with a filter included in the transmitter that filters the signal.

### Claim 4 (depends on 1)

The computer-implemented method of claim 1, wherein determining the one or more values for the one or more parameters comprises:
setting the one or more parameters to one or more predetermined values;
performing a first grid search using a first grid size associated with the one or more predetermined values to select one or more initial values for the one or more parameters; and
performing a second grid search that applies a second grid size to a vicinity of the one or more initial values to select one or more final values for the one or more parameters, wherein the second grid size is smaller than the first grid size.

### Claim 5 (depends on 1)

The computer-implemented method of claim 1, wherein determining the one or more values for the one or more parameters comprises:
performing one or more hill climbing operations to generate a set of locally optimized values associated with the one or more parameters; and
displacing one or more values included in the set of locally optimized values to generate a set of mutated values.