Abstract: Lack of expert resources and uncertain time to market are barriers to adopting machine learning by enterprise. AutoML holds the promise of lowering these barriers, yet it is mostly associated with computer vision. Erez Sali and Hila Lamm explain how to apply AutoML to supervised tasks with structured data and provide real-life examples and Kaggle benchmarks using AutoML for enterprise use cases.
Overloaded data science teams and lack of certainty in data science project results are slowing down the proliferation of machine learning in the realm of business. By using AutoML tools, data science teams can not only expedite their processes, but they can also widen their range of expertise, define predictive models for more directly-focused initiatives, and surpass time-to-delivery goals by a significant margin.
Using real-life cases, this session walks you through specific considerations of AutoML for the enterprise including:
* Enterprise data - from big data to very small training sets
* Automating preprocessing
* Time to delivery vs. model performance
* Comprehensibility - dealing with explaining a model built automatically
For each of the above topics Erez Sali will review the research background and data science practice, while Hila Lamm will provide the business stakeholder point of view.
Participants in this session will come away with an understanding of the key components of AutoML and how it enables enterprises to realize the full potential of an in-house data science team. You will leave this talk able to define specific challenges within your enterprise that can be modelled in order to advance business goals.