Menu
Another Industry First- On-Chain Neural Net Machine Learning Contracts Coming to the GNY Wallet
Updated: Feb 25
Quick Takeaways:
- The two latest GNY on-chain contracts mimic the way the human brain operates. The first to be deployed will be GNY’s Decentralized Word Vector Neural Net.
- Its learning framework will use a multilayer perceptron and embedding to build a profile and understanding of the interactions between users and products.
- The neural net will have the ability to optimize content suggestions for online news readers while creating a self-learning recommendation engine that responds to an individual’s preferences over time.
- Another potential use case concerns how health services could optimize public health advice to patients based on their confidential history, personalized health needs, and libraries of associated material, resulting in an improvement of health metrics.
Following the launch of our Mainnet in Q1 2021, the GNY Team has an ambitious agenda for introducing additional Machine Learning contracts to our Wallet. The next two on-chain contracts are both neural nets, and they represent yet another technical first for our blockchain team. Neural nets or neural networks are a series of algorithms that endeavor to recognize underlying relationships in a set of data through a process that mimics the way the human brain operates. Bringing two such powerful neural nets to our users is just a taste of the Machine learning capacity that the GNY Team will add to the platform by the year’s end and beyond.
To start, let’s take a look at the first neural net being added: The GNY Decentralized Word Vector Neural Net. This is a neural network based collaborative learning framework written in Javascript using TensorFlow.js and Node.js. that will use a multilayer perceptron to learn user-item interaction functions. GNY uses embedding to build profiles and understanding of the interactions between users and products. To do this, we leverage existing data of products, users, and ratings given by those users.
This embedding space helps the neural network better understand the interaction between products and users, and we can leverage this knowledge, combined with the user ratings of each product, to train a neural network. This is a classic regression approach, where the input is the learned embedding of products-user interaction, and the target/labels are product ratings given by the users. Using Multilayer Perceptron (MLP) to learn user-item interactions is an upgrade over matrix factorization, which is the most used variation of collaborative filtering. MLP can learn ANY continuous function and has high level of nonlinearities due to multiple layers making it well suited to learn user-item interaction function.
To Illustrate some potential ways this tool can provide value to organizations we have taken two GNY use case suggestions from community members and expanded them with some hypothetical clients.
Hypothetical Client # 1 – Apple News Stories.
The dilemma – How can Apple suggest the most relevant items to existing customers?
The solution – Using GNY’s Word Vector Neural Net, Apple could optimize page/content suggestions for users that visit their news portal on iPhone and iPad.
How it works – By analyzing each news story and creating word vectors, Apple can instantly make sure that it doesn’t suggest duplicate stories while creating a self-learning recommendation engine that responds to the preferences of individual readers over time.
Potential Results – Based on the publishing use case we featured in our white paper, Apple could see viewership increase up to 100% after several months.
Hypothetical Client # 2 – National Health Service Personal Care Portal
The dilemma – Medical records often hold under-utilized data that may help patients make healthier choices and aid professionals in pinpointing critical updates.
The solution – The GNY Word Vector Neural Net could optimize public health information, articles and tutorials to patients based on their confidential history, personalized health needs, and libraries of associated material.
How it works – The GNY Word Vector Neural Net would read the history of a patient’s care, treatment, and results to create a profile for this individual. Simultaneously, the neural net is interfacing with all available and approved wellness documentation. Based on these profiles and the user’s viewership history GNY then recommends next best health articles and information.
Potential Results – A decrease in medical visits and an improvement of health metrics (ex. blood pressure, weight, etc.) as well as improved leveraging and impact of public health materials.
Let us know in our Telegram chat group what you think of our potential use cases for GNY’s Decentralized Word Vector Neural Net. Please feel free to add your ideas for other areas you believe will benefit from this on-chain contract.
