Quantification and Mitigation of Epidemic using Deep-Learning on Distributed Peer to Peer Network- 1 min
this project is not Implemented yet its just one of my cool ideas that I wish to work on!!
Contagious Diseases are the deadliest killers, and there exists no pre-eminent approaches to predict, quantify and mitigate them. Existing mathematical models miserably fail on the real world due to their rigidity towards a specific kind of diseases and poor data collection. The scale of problem vs the organizational working capacity creates a bottleneck problem. The intensity of the data collections requirement may lead to privacy and data breach problems. I propose a novel solution to this problem by using Neural Networks on Ethereum Block Chain. Neural Networks are excellent at dimensionality reduction and supervised prediction which makes the model more flexible. First, the data is uploaded in real-time through a thin client of web3.js to the IPFS hash tables, the generated hash addresses are updated in the smart contract of the blockchain, this smart contract consists a pre-defined neural network and the generated hash address of the training data. The blockchain miners use this data to train a neural network as a ‘proof of work’ and the results will again be stored in the IPFS hash tables. The user can query the blockchain to get the IPFS hash address where the results of the neural network have been stored and in return will pay crypto-tokens (currently ether) to incentivize the miner and data uploader. This leads to a completely autonomous and immutable system for epidemic management. It is elastic as the system grows with the increase in problems, and secure as the whole system works on zero trusts and only cryptographic security.