DeAI - "Decentralized Artificial Intelligence" is a framework for hosting and training machine learning models on a distributed blockchain infrastructure. It requires only a single deployment of a machine learning model on an ethereum based smart contract, which can then be used to train or query predictions. Anyone, whether the person who deployed the model or another participant, can contribute data to train the model. This type of decentralized architecture can help to improve the accuracy of machine learning models with the help of computations locally over the data and if accuracy is improved, model is updated over the blockchain, thus privacy of data is maintained and at the same time it helps to improves the model. Users can access smart contracts to get model predictions. Users can give data to the smartcontract for winning incentives(POAP). if the Data supplied by the user is good then the model inside the smart contract is trained including that data. After the model has been retrained it is pushed onto the the Textile bucket whose IPNS address is public and developers all across the world can make use of our model.

DeAI showcase

How it's made

First of all Smart contracts are written using solidity and truffle which has contracts for giving prediction results using the stored ML models(perceptron model currently used). Another contract receives data from the user and trains the model if the accuracy of the model is improved then the data supplied was good data and the smart contract gives out a positive reply to the POAP module of the client app to trigger the incentive mechanism. Also if the model is improved by the data provided it is pushed to a textile bucket whose link is public. The client app is made using flutter framework which is then interfaced with the smart contracts and the textile bucket.

Technologies used