Trustworthy Deep Learning
In this session, attendees will learn about trustworthy deep learning, a new paradigm that involves designing and training neural networks that can make correct decisions with very high probability (i.e., near zero false positives). The paradigm will be illustrated through the example of deep learning startup Entrupy, whose system distinguishes between fake and real physical goods. Presented by Courant Institute Associate Professor and Chief Scientist of Entrupy, Lakshminarayanan “Lakshmi” Subramanian, the session will discuss the development process, including how the team set and achieved its trustworthy classification goals -- in this case, enabling a near zero probability of misclassifying a fake good as an authentic as well as maximizing the probability that an authentic good is accurately classified (i.e., minimizing false negatives), within the bounds of market expectations. He will also share how an array of different deep learning algorithms were trained on different regions of a physical item to increase accuracy of the classification engine, ultimately making it capable of separating genuine versions of a product from even the most high-quality counterfeit ones, often referred to as “super fakes.” Lastly, he will talk about the process of productizing the solution through an end-to-end system comprising a mobile device, portable hardware and a cloud-based object verification ecosystem.
Lakshminarayanan “Lakshmi” Subramanian is an Associate Professor of the Courant Institute of Mathematical Sciences at NYU and a co-founder of Entrupy, a startup that uses computer vision algorithms and microscopy to bring trust to transactions of high-value physical goods. He is a frequent speaker at universities and conferences and is the recipient of several awards including the NSF CAREER Award, faculty awards from Facebook, Google, IBM, Microsoft, Cisco and the C.V. Ramamoorthy Award. At NYU, Lakshmi leads the Open Networks and Big Data lab, conducting fundamental and applied research in open networking and big data analytics, with a particular focus on areas with direct applications to large-scale societal and developmental problems. Lakshmi is a co-founder of the Center for Technology and Economic Development (CTED) at NYU, which brings together students from several disciplines (CS, economics, healthcare, education, policy), and several technological of his technological innovations have been implemented in developing countries around the world.