{ immsrini }

About

I’m associated with Market Learning Labs various technical and founding roles.

MLLabs is an Deep Learning for enterprises company that is leveraging the power of AI to bridge the gap between perception and context while enabling easy-to-use applications. The fundamental principle, is that, we exploit the interplay of physical sciences and analytics.

Research interest: Statistics | Machine Learning | Cognitive Science

Understand the mechanism by which the brain translates environmental signals into intelligent behavior. I’m inquisitive about the working of our brain, its cognitive architecture and would like to build applications using artificial intelligence that mimic the nous. I’m particularly interested in Machine/Deep Learning and its applications. Theoretical aspects of Statistics like probability modeling and large scale statistical estimation problems intrigue me.

View { Publications and Patents | CV }

I’m an ardent classical music (the Indian flavor) listener.
I’m actively learning of Yajurveda (यजुर्वेदः).
I organize of the Deep Learning Bangalore (DLBLR) Meetup.

Blog

This post is the second in the series of understanding deep learning. To come up to speed with the current discussion, read the previous post. ##Interpretations of representation learning Deep learning has established that feature selection and engineering is not always useful. What do we replace future engineering with, in a representation learning framework? The objective is to learn a hierarchy of features which takes the output from the previous level as its input. The input to the next consequent level is the output of the previous level. Traditional algorithms like SVM and regressions are considered as “shallow” representations since these methods directly carry out classification from input data and do not have intermediate layers. This is where this type of representation learning gets its name “deep.”…