Talk on Visualizing Machine Learning

There is a lot of buzz around advanced analytics today, with a great deal of euphoria on AI applications and also a fair amount of hype around the sky-high expectations. At the other end of this spectrum there is deep discontent and disbelief on what the algorithms can achieve.

Ethical dilemmas aside, what lies at the root of this disconnect is a fundamental lack of understanding of machine learning algorithms. This underlying challenge with technology awareness leads to trust issues and finally manifests as stumbling blocks in adoption of the machine intelligence.

This was the topic of my talk last week in the monthly meeting organized by the New Jersey Data Science group. The session was titled “Visualising Machine Learning: Humanising the Intelligence”, and here is the deck.

I attempted to highlight this disconnect using live case studies from our work at Gramener, and illustrated the classic dilemma of black-box models with high accuracy but low adoption. Rest of the session touched upon the steps towards a possible resolution of this challenge by adopting a 3-stage visual solution framework, along with illustrations from public & project examples.