A blockchain framework for hosting and collaboratively training publicly available machine learning models with free predictions.
Sharing Updatable Models (SUM) on Blockchain is a framework that enables the hosting and collaborative training of machine learning models on a blockchain. It allows users to freely query models for predictions while contributing data through a validated process to improve model accuracy over time. The system uses smart contracts to manage data storage, model updates, and incentive mechanisms to ensure quality contributions.
Researchers and developers working on decentralized AI, blockchain applications, or collaborative machine learning systems who need transparent, publicly accessible models. It's also suitable for those exploring incentive mechanisms for data curation in open ecosystems.
It provides a unique blend of blockchain transparency with machine learning, offering free model predictions and a structured way for the community to improve models collaboratively. Unlike traditional centralized ML services, it ensures data permanence and public access while mitigating spam through built-in incentive mechanisms.
Sharing Updatable Models (SUM) on Blockchain
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Anyone can query hosted models for predictions without cost, as highlighted in the key features, promoting accessibility.
Users contribute data via a three-step validation process with incentive mechanisms, enhancing model accuracy over time, as described in the architecture flow.
Data and metadata are stored on-chain, ensuring permanence and accessibility for future uses, a core feature noted in the README.
Includes mechanisms like deposits or karma points to deter spam and encourage quality contributions, detailed in the incentive mechanism step.
Currently supports only simple models like Perceptron; complex models require off-chain computation, which may compromise decentralization, as admitted in the FAQ.
Ethereum fees can be prohibitive for frequent updates, though the FAQ notes they are decreasing with proof of stake, adding operational overhead.
The three-step validation for data addition introduces delays, making it unsuitable for applications needing real-time training or rapid iterations.