President & CEO
Cognitive Computing and Emerging Analytics
This session will address the need for cognitive computing platforms that combine Semantic Reasoning (i.e, machines that can Sense, Think, Act and Learn®)with Computational Intelligence (i.e., advanced analytic techniques) starting with statistical techniques, moving to machine learning, and then discussing the advantages of newer innovations such as the Massive Dynamics’ Representational Learning Machine™ (RLM), a high-dimensional mathematics engine and one of the most sophisticated in the world. The RLM produces provably-optimal predictive functions that best model a data set. Attendees will learn how cognitive computing can allow companies and governments to derive meaning from data and make decisions, take actions and learn – harnessing the power of big data.
Stephen F. DeAngelis is a technology and manufacturing sector entrepreneur and patent holder with over 25 years of experience helping to pioneer the application of advanced cognitive computing technologies and applied mathematics to commercial industries and governmental agencies. He is President and CEO of the cognitive computing company Enterra Solutions, LLC, and President and CEO of the computational intelligence company Massive Dynamics, LLC. He is currently a Visiting Professional Executive in Cognitive Reasoning Platforms in the Department of Chemistry at Princeton University (2016). He has also been a Visiting Scientist at the Software Engineering Institute (SEI) at Carnegie Mellon University and a Visiting Scientist in the Mathematical and Computational Sciences Directorate and the Center for Advanced Technology at the Oak Ridge National Laboratory. Mr. DeAngelis has been recognized as one of Esquire magazine’s “Best and Brightest” honorees (“The Innovator” —December 2006); as one of the “Top Influencers in Big Data” (Forbes —2012), and he became a contributing member of Wired magazine’s Innovation Insights blog (2014). In addition, he is a member of the Board of Directors of the Dalai Lama Center for Ethics and Transformative Values at the Massachusetts Institute of Technology and the Founder and Chairman of The Project for STEM Competitiveness, a non-profit organization whose mission is to bring outcomes-measured STEM education to middle school and high school students through exciting and inspiring project-based learning. Areas of expertise: Cognitive Computing (Artificial Intelligence + Applied Mathematics), Supply Chain, Digital Transformation, Analytics and Insights.
Democratizing Optimization: The Role of Algorithm Marketplaces in Analytics
Satalia (NPComplete Ltd) provides ‘Optimisation-as-a-Service’ to academia and industry and is a spin-out from the UK’s premier University, UCL. Satalia was recognises as a Gartner Cool Vendor in Data Science in 2016, the only UK company from five chosen globally. The Satalia SolveEngine was developed as a one-stop-shop for state-of-the-art optimisation algorithms giving industry immediate access to cutting edge technology that enables more intelligent use of compute resources and produces smarter, faster solutions to complex problems. Satalia provides new economic models to allow academics to commercialise their algorithmic innovations and get access to critical data to improve their research, and gives producers of software tools a new revenue mechanism to support the development and exposure of their software. This software platform includes five main components 1) a portal and API to allow the submission of optimisation problems and retrieve the solutions, 2) new optimisation translators and encoders to provide interoperability between algorithmic formats (such as MPS, SAT, CSP, FZN, LP, AMPL, SMT, OPB, etc), 3) machine learning to intelligently match the right algorithms to the problems it receives, 4) distributed computation, running on public or private clouds and HPC, and 5) a management console to allow academics to include their algorithms on the platform, commercialise their IP and gain access to valuable data to enable better research. The ‘Optimisation-as-a-Service’ platform is a conduit for academic algorithms into industry, and has wide reaching commercial applications across every industry sectors. Optimisation problems exist across the entire spectrum of computer science and engineering, and Satalia’s SolveEngine has already been used to solve hard computational problems across the world.
Dr. Daniel Hulme is the CEO of Satalia (NPComplete Ltd), that provides Artificial Intelligence solutions to solve industries hardest problems. Satalia’s core product is the SolveEngine, an Optimisation-as-a-Service API and web-portal that provides a conduit for cutting-edge optimisation algorithms to be easily and cheaply used by industry. Daniel has a Masters and Doctorate in AI from UCL, and is Director of UCL’s Business Analytics MSc in one of the world’s leading Computer Science departments; applying AI to solve business/social problems. He lectures in Computer Science and Business at undergraduate and postgraduate levels. Daniel has Advisory and Executive positions in many companies, he holds an international Kauffman Global Entrepreneur Scholarship and actively promotes entrepreneurship and technology innovation across the globe.
Senior Data Scientist, Analytics Server R&D
A Deep Learning Tutorial
There has recently been a surge of improvements in machine learning due to many innovations in deep learning field. Deep neural networks can learn to represent increasingly complex relationships in data using multiple layers of abstractions in a hierarchical manner. It is a very general way to model a problem and this generality proved to be very powerful in solving various challenging questions in machine learning and artificial intelligence. Deep learning methods have been pushing the state of the art for computer vision, natural language processing and other cognitive tasks as well as have been put use in various industries such as health care, customer intelligence, retail, finance and cyber security.
In this talk we will introduce the deep learning technique to ORMS practitioners in a holistic way. Goals of this tutorial talk are as follows:
1. Expose the audience in general to machine learning and in particular to deep learning approach.
2. Provide necessary background and terminology to be able to engage in discussions with researchers and practitioners in deep learning field.
3. Summarize various deep learning techniques that apply to different classes of problems.
4. Provide example solutions for representative problems.5. Provide pointers to different deep learning software
We aim to cover the following topics:
- A biologically inspired introduction to artificial neural networks as hierarchical function approximations.
- Formulating neural network training as an optimization problem and backpropagation algorithm.
- Training neural networks, the experience, regularization techniques for generalization.
- Stochastic gradient method to solve neural network training, variants and parallelization techniques.
- Convolutional neural networks and computer vision applications.
- Recurrent neural networks for sequential data, applications to natural language processing and time series forecasting.
- Autoencoders and unsupervised representation learning.
Mustafa is an Operations Research expert who likes to view the human activities as processes that can be modelled mathematically and optimized. During his PhD he worked on game theory models of supply chains selling to strategic customers. Earlier in his career at SAS he developed distributed large scale integer optimization algorithms for Marketing Optimization problems. He recently hopped on the Machine Learning, AI train and converted into a data scientist. Currently he sits at the intersection of theory, practice and development for SAS’s next generation Deep Learning and Cognitive Computing Tools. He is fascinated by all the recent successes of these approaches and eager to incorporate them into SAS’s offerings. As an optimization enthusiast he always looks into ways to improve the algorithms. Nowadays his favorites are the Distributed Stochastic Gradient methods. Mustafa is a father of two beautiful girls and is married to Ilknur, a Computer Science Evangelist, who leads a team of Machine Learning Scientists at SAS.
Head of Marketplace Optimization Advanced Development, Uber
and Paul M. Montrone Professor, Columbia University
A dominant trend of the past half-century is the global rise of market economics. Witness the freeing up of many previously centralized economies (China), expansion of free trade agreements (NAFTA) and unrestricted flows of capital and labor (EU). Technology has accelerated this trend at the micro level, enabling industries such as advertising (Google), transportation (Uber), lodging (Airbnb), finance (LendingClub), and retailing (Amazon), to organize not as traditional firms but as marketplaces – ecosystems that approach the perfect market ideal of full information, costless transactions and fully rational cognition. In this talk, I argue that building and growing these new marketplace businesses requires a new breed of marketplace engineers – professionals who combine deep understanding of technology, economics and data science. I discuss historical and recent examples of this trend and explore the implications for the field of operations research.
Garrett van Ryzin is the Paul M. Montrone Professor of Decision, Risk and Operations at Columbia University Business School. His research focuses on algorithmic pricing. Garrett has extensive consulting experience with leading firms in both established industries and technology startups. Since 2015, he has been Head of Marketplace Optimization Advanced Development at Uber Technologies. He is coauthor of the leading scientific book on revenue management, The Theory and Practice of Revenue Management, which won the 2005 Lanchester prize for best published work in operations research. He is an INFORMS and MSOM. Garrett received the B.S.E.E. degree from Columbia University, and the degrees of S.M. in Electrical Engineering and Computer Science and Ph.D. in Operations Research from MIT.
KDD Career Development Professor, Communications & Technology
Sloan School of Management
Finding Online Extremists in Social Networks
Online extremists in social networks pose a new form of threat to the general public. These extremists range from cyberbullies who harass innocent users to terrorist organizations such as ISIS that use social networks to spread propaganda. Currently social networks suspend the accounts of such extremists in response to user complaints, but these extremist users simply create new accounts and continue their activities. In this talk we present a new set of operational capabilities to help authorities mitigate the threat posed by online extremist groups in social networks. Using data from several hundred thousand extremist accounts on Twitter, we develop a behavioral model for these users, in particular what their accounts look like and who they connect with. This model is used to identify new extremist accounts by predicting if they will be suspended for extremist activity. We also use this model to track existing extremist users as they create new accounts by identifying if two accounts belong to the same user. Finally, we use this model as the basis for an efficient policy to search the social network for suspended users’ new accounts. Our search approach is based on a variant of the classic Polya’s urn setup. We find a simple characterization of the optimal search policy for this model under fairly general conditions. Our search policy and main theoretical results generalize easily to search problems in other fields.
Tauhid Zaman is the KDD Career Development Professor in Communications and Technology MIT Sloan School of Management. He received his BS, MEng, and PhD degrees in electrical engineering and computer science from MIT. His research interest is in behavioral analytics, with a focus on solving operational problems using behavioral models, modern statistical methods, and network algorithms. His work has been featured in Wired, Mashable, the LA Times, and Time Magazine.