Aggregated machine learning verification for database

A method, comprising:

• receiving, by a training aggregator in a blockchain network, a gradient calculation corresponding to a training iteration for machine learning model training from each training client of a plurality of training clients in the blockchain network, wherein each gradient calculation is calculated using a training dataset of a corresponding training client, wherein the training dataset and parameters for the machine learning model training are stored outside of a blockchain of the blockchain network, and wherein the each gradient calculation is stored in the blockchain;
• generating, by the training aggregator, a plurality of transaction proposals, each transaction proposal corresponding to a training client of the plurality of training clients and each transaction proposal containing the corresponding gradient calculation, a batch comprising a subset of samples from the dataset, a loss function identifying a cost metric, and an original model parameter, wherein the batch and the original model parameter are stored outside of the blockchain, and wherein the corresponding gradient calculation and the loss function are stored in the blockchain;
• receiving, by an endorser node in the blockchain network, the plurality of transaction proposals from the training aggregator;
• endorsing, by the endorser node, a transaction proposal, of the plurality of transaction proposals, in response to validating the corresponding gradient calculation using a verify gradient smart contract of the endorser node;
• and
• sending, by the endorser node, the endorsed transaction proposal to the training aggregator to generate a blockchain transaction using the endorsed transaction proposal.

A non-transitory computer readable medium comprising one or more instructions that when executed by one or more processors cause the one or more processors to perform:

• receiving, by a training aggregator in a blockchain network, a gradient calculation corresponding to a training iteration for machine learning model training from each training client of a plurality of training clients in the blockchain network, wherein each gradient calculation is calculated using a training dataset of a corresponding training client, wherein the training dataset and parameters for the machine learning model training are stored outside of a blockchain of the blockchain network, and wherein the each gradient calculation is stored in the blockchain;
• generating, by the training aggregator, a plurality of transaction proposals, each transaction proposal corresponding to a training client of the plurality of training clients and each transaction proposal containing the corresponding gradient calculation, a batch comprising a subset of samples from the dataset, a loss function identifying a cost metric, and an original model parameter, wherein the batch and the original model parameter are stored outside of the blockchain, and wherein the corresponding gradient calculation and the loss function are stored in the blockchain;
• receiving, by an endorser node in the blockchain network, the plurality of transaction proposals from the training aggregator;
• endorsing, by the endorser node, a transaction proposal, of the plurality of transaction proposals, in response to validating the corresponding gradient calculation using a verify gradient smart contract of the endorser node;
• and
• sending, by the endorser node, the endorsed transaction proposal to the training aggregator to generate a blockchain transaction using the endorsed transaction proposal.