Artificial Intelligence and Deep Learning for Decision Making with Large-scale Transactional Data
In this talk, I shall look at AI solutions over another type of signals, transactional data generated from our social and commercial world. A difference is that transactional data is usually heterogeneous, coming from diverse sources, and categorical in nature (representing attributes such as a person’s gender, marital status, hometown, career, or the types of movies they like). The correlations of those attributes are not known a priori. In MediaGamma, we have developed an audience decision engine (ADE) that can teach machines to make various predictions and then subsequently make optimal actions upon those predictions over terabytes transaction data. My talk will be focused on the ADE as an example to illustrate 1) how supervised and unsupervised learning can be developed to decipher the correlations among those attributes and use them to make various accurate predictions and forecasts, 2) how the knowledge can be transferred among those predictive tasks, and 3) how machines can pick up a signal and take an optimal decision over time against a pre-specific utility.
Dr. Jun Wang is Reader (Associated Professor) in Computer Science, University College London, and Director of MSc Web Science and Big Data Analytics. He is also Co-founder and Chief Scientist in MediaGamma, a UCL spin-out focusing on AI for intelligent audience decision making. His main research interests are in the areas of intelligent information systems, covering information retrieval, data mining, online advertising and deep learning. His team won the first global real-time bidding algorithm contest with 80+ participants worldwide. Jun has published over 100 research papers and is a winner of multiple “Best Paper” awards in information retrieval and data mining. He was a recipient of the Beyond Search – Semantic Computing and Internet Economics award sponsored by Microsoft Research and also received Yahoo! FREP Faculty award. MediaGamma has received the UCLB One-to- Watch award 2016.