Reasoning: The Next Frontier in Data Science
Submitted by Shailesh Kumar (@shkumar) on Thursday, 21 July 2016
The “Prediction Paradigm” in data science has come a long way. Today, we can build reasonably accurate models for complex prediction problems such as detecting objects in Images, answering Jeopardy questions, translating documents from one language to another, or recognising people from face images.
In this talk we will explore the next paradigm in data science - the “Reasoning Paradigm” that tries to optimize a “sequence of actions” leading from a “start state” to an “end state”. Prescribing a treatment plan for a set of symptoms, learning strategies for playing Chess or Go, solving multi-step problems in mathematics, maximizing life-time-value of a customer, having a goal driven conversation with a chat-bot, or connecting the dots on a knowledge graph are different flavours of multi-step reasoning problems that cannot be solved by the single-step prediction paradigm.
This talk will focus on two specific reasoning paradigms - Mathematical Reasoning and Reasoning over Knowledge Graphs. We will explore the building blocks for an intelligent reasoning engine that “explores” the space of possible solutions, “discovers” one or more solutions, characterizes the “quality” of each solution, “generalizes” to “similar” reasoning problems, and most importantly “learns” how to generate “better” solutions “faster” with practice - the holy grail of AI.
Shailesh Kumar is Chief Scientist and Co-Founder at ThirdLeap. He has 14 years over fifteen years of experience in applying and innovating machine learning, statistical pattern recognition, and data mining algorithms to hard prediction problems in a wide variety of domains including: remote sensing, text mining, bio-informatics, computer vision and image understanding, transaction data mining, retail analytics, neurological data, risk analytics in financial domain,and web analytics.