4Paradigm Won the First Transfer Learning Algorithm Contest Champion


The ‘Haoxin Cup’ big data algorithm competition, hosted by Qian Hai Zheng Xin limited and organized by Kesci.com came to an end. The competition focused on the topic of ‘transferlearning’. The dandange team, which had only one member, Luo Yuanfei, a 4Paradigm machine learning engineer, used a distributed parallel computing framework GDBT independently developed by 4Paradigm, and achieves excellent model effects while significantly reducing the workload of feature engineering. In the final interview, he showed a wealth of knowledge with accurate algorithm and stable field performance and won the championship.

The competition lasted two months and attracted more than 600 entrants across 242 teams. The competition received submissions of 145 model works and 17 algorithm schemes, and 10 teams were shortlisted in the final.

This algorithm competition selected transfer learning as its theme explicitly due to its great potential. Contestants were given 40,000 business credit loan data points and 4,000 cash loan data points, and then had to establish a credit score model for the cash loan business by transferring knowledge acquired from the credit loan data.