What you cannot do with Machine Learning
Section: Crisp Talk Technical level: Beginner
During this “boom” of machine learning and data driven technologies, there is an underlying belief that given enough data any problem is solvable. But like any other technology, machine learning is a tool, appropriate for some problems and not so appropriate for others. Though this talk I would like to remind the community of the things which cannot be done through machine learning.
- data taken without context can be worse than useless
- complex systems can fool you
- the metrics you use can have unintended consequences
Harsh Gupta is an engineer at Nilenso, India’s first employee owned software cooperative. His background is in mathematics and computing and he has spent a fair amount of time thinking about aspects of fairness in machine learning as part of his master’s thesis.
- “Community is more important than code”, Keynote at Kharagpur Open Source Summit 2017 https://www.youtube.com/watch?v=DZMejOBWpNM
- “FAT ML (Fairness, Accountability, and Transparency in Machine Learning) for Lawyers and Lawmakers”, CISxScholars, New Delhi, India, https://www.dropbox.com/s/j0q40dkei2qrec5/CIS_talk_presentation.pdf?dl=0
- “Tech is WEIRD”, lightning talk at SciPy 2016, Austin https://www.youtube.com/watch?v=sv9S-25XKe4
- “What’s new with sympy solvers”, lightning talk at SciPy 2015, Austin https://youtu.be/YCxQI4C34j8?t=7m55s)
A novel Interactive Framework for semi-automated labeling when ground truth resides in free text
In any multi-class supervised learning problem, labeling of training examples is imperative. In most cases, we take expert help in order to execute the annotation, which is time-consuming and often inconsistent. In this talk, we will explain an interactive topic modeling framework to label training examples where the ground truth resides in free text. They key takeaways of this talk will be 1) A … more