•  
• Tweet
•  
•        Print
•  
•         Track Keywords
•  
• 

Bailey McCann, Opalesque New York

Jürgen Schmidhuber is director of the Swiss Artificial Intelligence Lab and professor of artificial intelligence at the University of Lugano, Switzerland. He is an expert on machine learning, financial data prediction, and mathematically optimal universal Artificial Intelligence (AI) among other topics.

Schmidhuber was recently interviewed by Sona Blessing for Opalesque Radio.

His expertise gives him insights into the use of artificial intelligence for investing in alternatives. He explains that, research into the use of neural networks allows algorithms to become sufficiently predictive of market conditions only when they are simple.

"Many people are using artificial neural networks and training them on a set of past stock market data. They try to reproduce the previous data and give the network lots of examples in order to train it. At some point, from this you get to a good predictive model on the training set. So then the question becomes how good is this network with unseen, future data? How can you generalize from past stock market data to the new unknown data coming in the future? What many people are trying to do is train their networks based on all of the points of a given data set. But the trianing should also be simple. There should be no complexity, and a low description size."

According to Schmidhuber, even with predictive algorithms......................

To view our full article Click here