# US 2025/0147664 A1 · Tracking read voltages in memory devices using deep neural networks

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

- Number: US 2025/0147664 A1
- Document ID: US20250147664A1
- Status: Published application
- Kind: A1
- Application number: 18/500,524
- Filed: 2023-11-02
- Published: 2025-05-08
- Earliest claimed: 2023-11-02
- Assignees: SK hynix Inc.
- Applicants: SK hynix Inc.
- Inventors: Fan Zhang; Meysam Asadi; Haobo Wang
- Web page: https://patentgrep.com/patent/US20250147664A1

## CPC

- G06F 3/0611: {in relation to response time}
- G06F 3/0653: {Monitoring storage devices or systems}
- G06F 3/0679: {Non-volatile semiconductor memory device, e.g. flash memory, one time programmable memory [OTP]}
- G06N 3/048: Activation functions
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Devices, systems, and methods for improving performance of a memory device are described. An example method includes extracting parameters, which include a read threshold set, from each of a first set of host reads, replacing, based on the parameters, at least one host read from a second set of host reads by at least one host read from the first set of host reads, using a deep neural network (DNN) to generate an updated read threshold set, wherein an input to the DNN comprises the parameters from each of the second set of host reads subsequent to the replacing, and applying the updated read threshold set to the memory device to retrieve information from the memory device. In an example, the number of the first set of host reads is at least two orders of magnitude greater than the number of the second set of host reads.

## Claims

### Claim 1

A method for improving performance of a memory device, comprising:
extracting one or more parameters from each of a first plurality of host reads, wherein the one or more parameters comprises a read threshold set;
replacing, based on the one or more parameters, at least one host read from a second plurality of host reads by at least one host read from the first plurality of host reads, wherein a number of the first plurality of host reads is at least two orders of magnitude greater than a number of the second plurality of host reads;
using a deep neural network (DNN) to generate an updated read threshold set, wherein an input to the DNN comprises the one or more parameters from each of the second plurality of host reads subsequent to the replacing; and
applying the updated read threshold set to the memory device to retrieve information from the memory device.

### Claim 2 (depends on 1)

The method of claim 1, wherein the one or more parameters comprises at least one of a checksum, an asymmetric ratio, a ones count, or a zeroes count.

### Claim 3 (depends on 1)

The method of claim 1, wherein the replacing the at least one host read comprises:
determining that a metric associated with the one or more parameters of the at least one host read from the first plurality of host reads is greater than the metric associated with the one or more parameters of the at least one host read from the second plurality of host reads.

### Claim 4 (depends on 3)

The method of claim 3, wherein the metric associated with the one or more parameters of the at least one host read from the second plurality of host reads is less than each metric associated with the one or more parameters of other corresponding host reads from the second plurality of host reads.

### Claim 5 (depends on 1)

The method of claim 1, wherein a history read threshold set is replaced by the updated read threshold set prior to applying the updated read threshold set.

### Claim 6 (depends on 1)

The method of claim 1, wherein the input to the DNN comprises the one or more parameters from a garbage collection (GC) read or a background media scan read.

### Claim 7 (depends on 1)

The method of claim 1, wherein the memory device is a non-volatile memory device, and wherein the first plurality of host reads are initiated by a host device that is communicatively coupled to the memory device.

### Claim 8

An apparatus for improving performance of a memory device, comprising:
a memory controller; and
a deep neural network (DNN) communicatively coupled to the memory controller,
wherein the memory controller is configured to:
extract one or more parameters from each of a first plurality of host reads, wherein the one or more parameters comprises a read threshold set;
replace, based on the one or more parameters, at least one host read from a second plurality of host reads by at least one host read from the first plurality of host reads, wherein a number of the first plurality of host reads is at least two orders of magnitude greater than a number of the second plurality of host reads;
use a deep neural network (DNN) to generate an updated read threshold set, wherein an input to the DNN comprises the one or more parameters from each of the second plurality of host reads subsequent to the replacing; and
apply the updated read threshold set to the memory device to retrieve information from the memory device.

### Claim 9 (depends on 8)

The apparatus of claim 8, wherein the one or more parameters comprises at least one of a checksum, an asymmetric ratio, a ones count, or a zeroes count.

### Claim 10 (depends on 8)

The apparatus of claim 8, wherein the memory controller is configured, as part of replacing the at least one host read, to:
determine that a metric associated with the one or more parameters of the at least one host read from the first plurality of host reads is greater than the metric associated with the one or more parameters of the at least one host read from the second plurality of host reads.

### Claim 11 (depends on 10)

The apparatus of claim 10, wherein the metric associated with the one or more parameters of the at least one host read from the second plurality of host reads is less than each metric associated with the one or more parameters of other corresponding host reads from the second plurality of host reads.

### Claim 12 (depends on 8)

The apparatus of claim 8, wherein a history read threshold set is replaced by the updated read threshold set prior to applying the updated read threshold set.

### Claim 13 (depends on 8)

The apparatus of claim 8, wherein the input to the DNN comprises the one or more parameters from a garbage collection (GC) read or a background media scan read.

### Claim 14 (depends on 8)

The apparatus of claim 8, wherein the memory device is a non-volatile memory device, and wherein the first plurality of host reads are initiated by a host device that is communicatively coupled to the memory device.

### Claim 15

A non-transitory computer-readable storage medium having instructions stored thereupon for improving performance of a memory device, comprising:
instructions for extracting one or more parameters from each of a first plurality of host reads, wherein the one or more parameters comprises a read threshold set;
instructions for replacing, based on the one or more parameters, at least one host read from a second plurality of host reads by at least one host read from the first plurality of host reads, wherein a number of the first plurality of host reads is at least two orders of magnitude greater than a number of the second plurality of host reads;
instructions for using a deep neural network (DNN) to generate an updated read threshold set, wherein an input to the DNN comprises the one or more parameters from each of the second plurality of host reads subsequent to the replacing; and
instructions for applying the updated read threshold set to the memory device to retrieve information from the memory device.

### Claim 16 (depends on 15)

The non-transitory computer-readable storage medium of claim 15, wherein the one or more parameters comprises at least one of a checksum, an asymmetric ratio, a ones count, or a zeroes count.

### Claim 17 (depends on 15)

The non-transitory computer-readable storage medium of claim 15, wherein the instructions for replacing the at least one host read comprises:
instructions for determining that a metric associated with the one or more parameters of the at least one host read from the first plurality of host reads is greater than the metric associated with the one or more parameters of the at least one host read from the second plurality of host reads, wherein the metric associated with the one or more parameters of the at least one host read from the second plurality of host reads is less than each metric associated with the one or more parameters of other corresponding host reads from the second plurality of host reads.

### Claim 18 (depends on 15)

The non-transitory computer-readable storage medium of claim 15, wherein a history read threshold set is replaced by the updated read threshold set prior to applying the updated read threshold set.

### Claim 19 (depends on 15)

The non-transitory computer-readable storage medium of claim 15, wherein the input to the DNN comprises the one or more parameters from a garbage collection (GC) read or a background media scan read.

### Claim 20 (depends on 15)

The non-transitory computer-readable storage medium of claim 15, wherein the memory device is a non-volatile memory device, and wherein the first plurality of host reads are initiated by a host device that is communicatively coupled to the memory device.
