Data Science is the study of the extraction of knowledge from data. What if we extract partial or inaccurate knowledge? This illusion of knowledge would lead us to make wrong decisions, with sometimes disastrous consequences such as in the case of medical diagnosis, security or other life and death situations. In this talk we’ll present ideas on how to validate the extracted knowledge by its predictive power and not by its ability to explain the past. We’ll also discuss special techniques for predicting very rare high value events.
Dr. Richard Robehr Bijjani has been developing practical machine learning products for security, biometric and personal health products for over 25 years. In 1990, Richard managed R&D during the development of a dynamic signature verification product at Kumahira Inc., one of the very first biometrics products in the industry. He later founded CompSec Biometrics to commercialize his invention for fingerprint bio-authentication. In 1994, he joined InVision Technologies as head of the Algorithm and Machine Vision group where he oversaw the development effort that led to the world’s first successful certification. He went on to design and certify multiple EDS systems for InVision (now Morpho Detection) and later for Vivid (now L3). In 2002, he co-founded Reveal Imaging (now an SAIC company) where he designed and developed the world’s highest performing automated explosive detection systems to date, which also happen to be the least expensive and the smallest. In January 2012, Richard founded Robehr Analytics which provides focused consulting services to key government agencies. In June 2012 he became an advisor for nCrypted Cloud which enables privacy, security and collaboration in the cloud. In October 2012 Richard co-founded Quanttus Inc. a consumer health care company that plans to forever change the way we view our personal health. Dr. Bijjani has a Ph.D. in Electrical Engineering from Rensselaer Polytechnic Institute.