# US 2020/0394471 A1 · Efficient database maching learning verification

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

- Number: US 2020/0394471 A1
- Document ID: US20200394471A1
- Status: Published application
- Kind: A1
- Application number: 16/438,500
- Filed: 2019-06-12
- Published: 2020-12-17
- Earliest claimed: 2019-06-12
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Venkata Sitaramagiridharganesh Ganapavarapu; Kanthi Sarpatwar; Karthikeyan Shanmugam; Roman Vaculin
- Web page: https://patentgrep.com/patent/US20200394471A1

## CPC

- G06N 20/00: Machine learning
- G06F 18/2148: {characterised by the process organisation or structure, e.g. boosting cascade}
- G06F 18/2185: {the supervisor being an automated module, e.g. intelligent oracle}
- G06N 3/08: Learning methods
- G06N 3/09: Supervised learning
- G06N 3/098: Distributed learning, e.g. federated learning
- G06N 7/08: using chaos models or non-linear system models
- G06Q 20/401: {Transaction verification}
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

An example operation may include one or more of generating, by a training participant client comprising a training dataset, a plurality of transaction proposals that each correspond to a training iteration for machine learning model training related to stochastic gradient descent, the machine learning model training comprising a plurality of training iterations, the transaction proposals comprising a gradient calculation performed by the training participant client, a batch from the private dataset, a loss function, and an original model parameter, receiving, by one or more endorser nodes of peers of a blockchain network, the plurality of transaction proposals, and evaluating each transaction proposal.

## Claims

### Claim 1

A system, comprising:
a training participant client, comprising a training dataset and configured to:
generate a plurality of transaction proposals that each correspond to a training iteration for machine learning model training related to stochastic gradient descent, the machine learning model training comprising a plurality of training iterations, the transaction proposals comprising a gradient calculation performed by the training participant client, a batch from the private dataset, a loss function, and an original model parameter; and
a blockchain network, comprising:
one or more endorser nodes or peers, each comprising a verify gradient smart contract configured to:
receive the plurality of transaction proposals; and
evaluate each transaction proposal.

### Claim 2 (depends on 1)

The system of claim 1, wherein the verify gradient smart contract evaluates the transaction proposals comprises the verify gradient smart contract is configured to:
evaluate the loss function at two different points;
calculate a difference between the loss functions at the two different points;
compare the difference to a product of a size of the batch, a random scalar, and a dot product between the gradient calculation and a random direction; and
determine if the comparison is greater than a predetermined threshold.

### Claim 3 (depends on 2)

The system of claim 2, wherein the predetermined threshold is application dependent.

### Claim 4 (depends on 2)

The system of claim 2, wherein the verify gradient smart contract evaluates the loss function at two different points comprises the verify gradient smart contract is further configured to:
generate the random scalar and the random direction; and
obtain a new model parameter by perturbing an original model parameter in the random direction with a step size equal to a value of the random scalar.

### Claim 5 (depends on 4)

The system of claim 4, wherein verify gradient smart contract calculates the difference between the loss functions at the two different points comprises the verify gradient smart contract is further configured to:
evaluate a difference in the loss function of all the samples in the batch between the new model parameter and the original model parameter.

### Claim 6 (depends on 5)

The system of claim 5, wherein in response to the comparison is greater than the predetermined threshold, the verify gradient smart contract approves the corresponding transaction proposal, wherein in response to the difference is not greater than the predetermined threshold, the verify gradient smart contract rejects the corresponding transaction proposal.

### Claim 7 (depends on 6)

The system of claim 6, wherein the verify gradient smart contract is further configured to:
provide endorsements for approved transaction proposals to the training participation client.

### Claim 8

A method, comprising:
generating, by a training participant client comprising a training dataset, a plurality of transaction proposals that each correspond to a training iteration for machine learning model training related to stochastic gradient descent, the machine learning model training comprising a plurality of training iterations, the transaction proposals comprising a gradient calculation performed by the training participant client, a batch from the private dataset, a loss function, and an original model parameter;
receiving, by one or more endorser nodes of peers of a blockchain network, the plurality of transaction proposals; and
evaluating each transaction proposal.

### Claim 9 (depends on 8)

The method of claim 8, wherein each of the one or more endorser nodes or peers comprising a verify gradient smart contract, wherein evaluating each transaction proposal comprises the verify gradient smart contract:
evaluating the loss function at two different points;
calculating a difference between the loss functions at the two different points;
comparing the difference to a product of a size of the batch, a random scalar, and a dot product between the gradient calculation and a random direction; and
determining if the comparison is greater than a predetermined threshold.

### Claim 10 (depends on 9)

The method of claim 9, wherein the predetermined threshold is application dependent.

### Claim 11 (depends on 9)

The method of claim 9, wherein evaluating the loss function at two different points comprises the verify gradient smart contract:
generating the random scalar and the random direction; and
obtaining a new model parameter by perturbing the original model parameter in the random direction with a step size equal to a value of the random scalar.

### Claim 12 (depends on 11)

The method of claim 11, wherein calculating the difference between the loss functions at the two different points comprises the verify gradient smart contract:
evaluating a difference in the loss function of all the samples in the batch between the new model parameter and the original model parameter.

### Claim 13 (depends on 12)

The method of claim 12, wherein in response to the comparison is greater than the predetermined threshold, the method further comprising:
approving, by the verify gradient smart contract, the corresponding transaction proposal,
wherein in response to the difference is not greater than the predetermined threshold, the method further comprising:
rejecting, by the verify gradient smart contract, the corresponding transaction proposal.

### Claim 14 (depends on 13)

The method of claim 13, further comprising:
providing endorsements, by the verify gradient smart contract, to the training participation client for approved transaction proposals.

### Claim 15

A non-transitory computer readable medium comprising instructions, that when read by a processor, cause the processor to perform:
generating, by a training participant client comprising a training dataset, a plurality of transaction proposals that each correspond to a training iteration for machine learning model training related to stochastic gradient descent, the machine learning model training comprising a plurality of training iterations, the transaction proposals comprising a gradient calculation performed by the training participant client, a batch from the private dataset, a loss function, and an original model parameter;
receiving, by one or more endorser nodes of peers of a blockchain network, the plurality of transaction proposals; and
evaluating each transaction proposal.

### Claim 16 (depends on 15)

The non-transitory computer readable medium of claim 15, wherein each of the one or more endorser nodes or peers comprising a verify gradient smart contract, wherein evaluating each transaction proposal comprises the verify gradient smart contract:
evaluating the loss function at two different points;
calculating a difference between the loss functions at the two different points;
comparing the difference to a product of a size of the batch, a random scalar, and a dot product between the gradient calculation and a random direction; and
determining if the comparison is greater than a predetermined threshold, wherein the predetermined threshold is application dependent.

### Claim 17 (depends on 16)

The non-transitory computer readable medium of claim 16, wherein evaluating the loss function at two different points comprising the verify gradient smart contract:
generating the random scalar and the random direction; and
obtaining a new model parameter by perturbing the original model parameter in the random direction with a step size equal to a value of the random scalar.

### Claim 18 (depends on 16)

The non-transitory computer readable medium of claim 16, wherein calculating the difference between the loss functions at the two different points comprises the verify gradient smart contract:
evaluating a difference in the loss function of all the samples in the batch between the new model parameter and the original model parameter.

### Claim 19 (depends on 18)

The non-transitory computer readable medium of claim 18, wherein in response to the comparison is greater than the predetermined threshold, the instructions cause the processor to further perform:
approving, by the verify gradient smart contract, the corresponding transaction proposal,
wherein in response to the comparison is not greater than the predetermined threshold, the instructions cause the processor to further perform:
rejecting, by the verify gradient smart contract, the corresponding transaction proposal.

### Claim 20 (depends on 19)

The non-transitory computer readable medium of claim 19, wherein the instructions cause the processor to further perform:
providing endorsements, by the verify gradient smart contract, to the training participation client for approved transaction proposals.
