This event has ended. Visit the official site or create your own event on Sched.
Tom Tom champions civic innovation, creativity, and entrepreneurship in America’s hometowns.

[Back to Tom Tom Festival]
Back To Schedule
Thursday, April 11 • 2:50pm - 3:20pm
Applying Machine Learning to Dynamic Operations

Log in to save this to your schedule, view media, leave feedback and see who's attending!

Feedback form is now closed.
There is significant rhetoric about Machine Learning, but less clarity on how it’s being applied to solve operational problems.  The majority of reinforcement-based machine learning systems continues to face limitations.  Those limitations include: Variance due to multivariate nature of the operations it’s learning, Latency due to the nature of the reinforcement learning model, Bias due to the manual feature engineering and tuning that happens.
We will first explore the challenges facing ML as it attempts to predict operational systems.  Most operational systems are non-deterministic so we will assume it is that type of system that we are trying to develop an advance predictive model of.  
Assume first that system is static, single variate, and we can use a learning representation to generate a probabilistic set of outcomes.  After repeated runs we would develop a model of the outcomes.  
Accurately representing that model so that it could be used to predict future outcomes would involve some type of tuning.  Current machine learning involves complex feature analysis and manual tuning over repeated runs to achieve a model that does not over fit or underfit, so that it can accurately predict future outcomes.
But most systems are multi-variate (have multiple variable inputs) that drive their various outcomes.  This multi-dimensionality factorially increases the complexity of the model we are trying to construct; as there are many combinations of values, across many combinations of inputs, that create the varying outcomes.  
Finally, most systems are dynamic rather than static.  So not only do the multivariate input values change dynamically, their correlation to the outcomes also changes dynamically.  This dynamic adds the requirement of continuous learning, in addition to that of continuous forecasting.
Is it possible to construct a system that can effectively learn and predict this kind of model?
This discussion will explore this problem of modeling dynamic Business Operations and the challenge it presents to Reinforcement Learning.  Then we will share practical examples of applying such types of machine learning to dynamic business operations, to improve operational outcomes.  In all cases, the goal of the system is to predict future events, such that people, equipment and processes can be optimized, and disruption minimized.

You need this ticket from Eventbrite to sign up: Applied Machine Learning Conference.

avatar for Tom Brock

Tom Brock

CEO, ConvergentAI
Tom brings over 30 years experience in software and analytics.  He has had roles ranging from COO, CIO, GM and Strategy, spanning from industrial automation to healthcare analytics.  He has BS degree in Electrical Engineering from Georgia Tech, is a graduate of GE’s management... Read More →


Thursday April 11, 2019 2:50pm - 3:20pm EDT
Violet Crown: Theater 4 200 W Main St, Charlottesville, VA 22902, USA