Abstract: Last over two months, hundreds of universities and students studied the implementation of Federated Learning application, collaboratively exploring the frontier fields of Fintech.


On August 26, 2019, Fintechathon – the first WeBank Fintech college technology competition (hereinafter referred to as “competition”) was successfully concluded in Shenzhen Qian Hai Vanke International Conference Center.

The competition lasts for two months and 11 days. Since the start of the competition, 178 teams and 659 participants have signed up, covering numerous top universities both domestically and internationally, as well as all levels of education background, from undergraduate to graduate.

The competition is based on the frontier of Fintech and consists of two parts: Artificial Intelligence and Blockchain. It is a stage for all domestic and international college students and is committed to assisting them to explore technological breakthroughs and application innovations in the frontier of Fintech.


Challenge the new direction of AI, college students study the implementation of Federated Learning application

After 30 hours of marathon programming, the champion was ” Federated We-car insurance – horizontal federated learning and 5G technology based personalized auto insurance pricing scheme” from “404 not found team”. It is reported that the two members of the champion team are from Peking University and the University of California, San Diego respectively. The winning production is based on federated learning and 5G technology, collecting data of users’ driving habits, building automobile insurance model and creating an innovative automobile insurance pricing scheme.

Besides, PPNE project submitted by members of Wuhan University, the FATE data platform (FDP) submitted by members of Northeastern University and The University of Sydney were awarded the second and third place respectively. The winning schemes were “Bank Transaction Network Association” and “Data Rights Trading Platform”, which inspired the thinking on the further application of Federated Learning from the perspective of practical financial scenarios. More importantly, the research on federated learning technology conducted by college students vitalizes the federated learning ecosystem.


The Federated Learning open-source platform FATE drives the value of AI

With the advance in science and technology, social demand is bringing forward new higher requirements for professional talent. As a leading technology in the digital era, federated learning is a key solution to data privacy and security. As a distributed machine learning paradigm, federated learning can effectively solve the problem of data silos, allowing participants to collaboratively build models without exchanging data, and thereby breaking data silos and realizing AI collaboration. The competition requires participating teams to design an application of the federated learning based on the Federated Learning open-source platform FATE, which is also intended to further promote the implementation of data security in the field of AI.

As the world’s first industrial-scale open-source framework for federated learning, Federated AI Technology Enabler (FATE) provides a secure computing framework to support federated learning algorithms. It implements secure computing protocols based on homomorphic encryption and multi-party computing, and supports secure computing of federated learning architecture and various machine learning algorithms. It includes classical ML algorithms such as logistic regression and gradient ascension tree, as well as cutting-edge researches such as deep learning and transfer learning.

By guiding college students to get familiar with and use the excellent industrial-scale open-source platform FATE, the competition reduces the threshold of cutting-edge technologies. The competition aims to provide both a demonstration and an exploration for the industry in solving the talent problem and to work with colleges and students to create a more excellent and expanding open-source ecosystem.