Tuesday, August 15, 2017

Machine Learning

(Translation of my Tamil blog at http://justexperience.blogspot.in/2017/08/blog-post_15.html)

"Meet me tomorrow after three.  My classes get over by then.  Let's talk about your project."

He hesitated.

"Why? Aren't you available?"

Suddenly his face cleared.  "No Sir, I'll come."

I know.  It's a Friday, a day when new movies release.  He must have quickly altered the plan for me.

I have a soft corner for Tamilvaanan.  He comes from a family of farmers.  He has told me about the hard physical work their family goes through for a good part of every year.  He believes education can make a difference to his family.  He isn't sharp but is capable of hard work.

He's my only hope in a set of 20 ME students.  Most students join masters because either they didn't get a job or found a suitable alliance for marriage.  Tamil was clear about learning.

He came on Friday.

"Do you know anything about  Astrology?"

"No Sir.  I don't believe in it."

It's a surprise for me who used Linda Goodman for flirting during college days.  Generation gap, I thought.

"Well, nor do I believe in Quantum Mechanics.  But studied it.  So, you also need to study Astrology.  Basics should do.  I'll arrange for it."

"You also need to focus on machine learning.  It's a core course for you this semester.  So, pay attention to it as well.  Your project will be an application of machine learning on Astrology."

"I don't have great beliefs in Astrology, but I have seen a few surprises.  Given a horoscope, my friend could predict the physiology, teeth structure etc.  We will go just that far.  We will validate a correlation of a few attributes of people to the planetary positions in their horoscope.  For example, we will take politicians as an example.  Their prominent attributes are oratory, charisma, hard work, street-smartness, diplomacy etc."

"We won't do definite prediction but will look at the probability.  We will build an unsupervised learning with the horoscopes and the attributes of the subjects and form a cluster.  We will test it with a few samples to determine the probability of those subjects to be successful politicians.  This should be a good one to complete in six months.  If possible, I'll extend it, with a Ph.D. student next year to determine if I'll get my dinner.  We won't go that far now."

"I've spoken to Balu, my cousin.  Spend four hours tomorrow and day after with him to get the basics.  Remember to collect the reference books.  He may offer them to you, but I'd rather buy it for us.  The university library would take a year to get those books."

When I met Tamil on Monday, he looked dull.

"Sir, can I do a different project in Machine Learning?"

I stared at him coldly.

"Sir, my uncle doesn't like it, Sir.  We are traditionally a rationalist family for three generations.  We don't go to temples or believe in Astrology."

I lost my cool.

"Idiot!  Don't you realize that we have an opportunity to analyze Astrology without subjective interpretation by Astrologers?  This is a good chance to make use of Machine Learning for impartial analysis.  You may even prove that Astrology is all fake."

"But Sir,... what if it's not?"

No comments: