Health Discovery Corporation (OTCBB: HDVY) announced today that Support Vector Machine technology was used to take first place in solving a complex marketing problem in the recent Active Learning Challenge. The results of the Active Learning Challenge were presented this past week at the 13 th International Conference on Artificial Intelligence and Statistics, in Sardinia, Italy.
The sponsors of the Active Learning Challenge, included the European Network of Excellence, Pascal2, the French Telecom Company Orange (Euronext: FTE, NYSE: FTE), Microsoft Corporation (NasdaqGS: MSFT), ETH Zurich, and the IEEE, the world’s largest professional association dedicated to advancing technological innovation and excellence for the benefit of humanity.
Teams from both industry and academia from all over the world participated in the challenge. Each team could choose any hardware, software and algorithms to solve the challenge problems. Some of the classifiers used in the Active Learning Challenge included linear classifiers, non-linear kernels, Naïve Bayes, Nearest Neighbors, Neural Networks, Bayesian Network, and Bayesian Neural Networks, Random Forests, and Support Vector Machines.
The winners of the marketing challenge included two students from the National Taiwan University, Ming-Hen Tsai and Chia-Hua Ho from the group lead by Professor Chih-Jen Lin, who successfully applied Support Vector Machines to the complex business marketing problem. The Taiwanese team finished second overall on all six challenge problems behind the team from Intel (NasdaqGS: INTC) and ahead of the team from CoreLogic (NYSE: CLGX) and eBay Inc. (NasdaqGS: EBAY).Students under Professor Lin’s tutelage have ranked among the top participants in several past challenges, including in the 2008 IEEE World Congress of Computation Intelligence Causality Challenge and the 2009 KDD Cup, applying Support Vector Machines to a wide variety of problems. This year the participants addressed a complex, machine learning problem of growing importance for all businesses: Building predictive models when massive amounts of unlabeled data are available and only few queries for labeled data can be placed.