Caste Brahmins, classification, and ANN

1. Caste Brahmins:

First, a clarification: No, I was not born in any one of the Brahmin castes, particularly, not at all in the Konkanastha Brahmins’ caste.

Second, a suggestion: Check out how many caste-Brahmins have made it to the top in the Indian and American IT industry, and what sort of money they have made—already.

No, really.

If you at all bother visiting this blog, then I do want you to take a very serious note of both these matters.

No. You don’t have to visit this blog. But, yes, if you are going to visit this blog, to repeat, I do want you to take  matters like these seriously.

Some time ago, perhaps a year ago or so, a certain caste-Brahmin in Pune from some place (but he didn’t reveal his shakha, sub-caste, gotra, pravar, etc.) had insulted me, while maintaining a perfectly cool demeanor for himself, saying how he had made so much more money than me. Point taken.

But my other caste-Brahmin “friends” kept quiet at that time; not a single soul from them interjected.

In my off-the-cuff replies, I didn’t raise this point (viz., why these other caste-Brahmins were keeping quiet), but I am sure that if I were to do that, then, their typical refrain would have been (Marathi) “tu kaa chiDatos evhaDa, to tar majene bolat hotaa.” … English translation: Why do you get so angry? He was just joking.

Note the usual caste-Brahmin trick: they skillfully insert an unjustified premise; here, that you are angry!

To be blind to the actual emotional states or reactions of the next person, if he comes from some other caste, is a caste-habit with the caste-Brahmins. The whole IT industry is full of them—whether here in India, or there in USA/UK/elsewhere.

And then, today, another Brahmin—a Konkanastha—insulted me. Knowing that I am single, he asked me if I for today had taken the charge of the kitchen, and then, proceeded to invite my father to a Ganesh Pooja—with all the outward signs of respect being duly shown to my father.


Well, coming back to the point which was really taken:

Why have caste-Brahmins made so much money—to the point that they in one generation have begun very casually insulting the “other” people, including people of my achievements?

Or has it been the case that the people of the Brahmin castes always were this third-class, in terms of their culturally induced convictions, but that we did not come to know of it from our childhood, because the elderly people around us kept such matters, such motivations, hidden from us? May be in the naive hope that we would thereby not get influenced in a bad manner? Possible.

And, of course, how come these caste-Brahmins have managed to attract as much money as they did (salaries in excess of Rs. 50 lakhs being averagely normal in Pune) even as I was consigned only to receive “attract” psychic attacks (mainly from abroad) and insults (mainly from those from this land) during the same time period?

Despite all my achievements?

Do take matters like these seriously, but, of course, as you must have gathered by now, that is not the only thing I would have, to talk about. And, the title of this post anyway makes this part amply clear.


2. The classification problem and the ANNs:

I have begun my studies of the artificial neural networks (ANNs for short). I have rapidly browsed through a lot of introductory articles (as also the beginning chapters of books) on the topic. (Yes, including those written by Indians who were born in the Brahmin castes.) I might have gone through 10+ such introductions. Many of these, I had browsed through a few years ago (I mean only the introductory parts). But this time round, of course, I picked them up for a more careful consideration.

And soon enough (i.e. over just the last 2–3 days), I realized that no one in the field (AI/ML) was talking about a good explanation of this question:

Why is it that the ANN really succeeds as well as it does, when it comes to the classification tasks, but not others?

If you are not familiar with Data Science, then let me note that it is known that ANN does not do well on all the AI tasks. It does well only on one kind of them, viz., the classification tasks. … Any time you mention the more general term Artificial Intelligence, the layman is likely to think of the ANN diagram. However, ANNs are just one type of a tool that the Data Scientist may use.

But the question here is this: why does the ANN do so well on these tasks?

I formulated this question, and then found an answer too, and I would sure like to share it with you (whether the answer I found is correct or not). However, before sharing my answer, I want you to give it a try.

It would be OK by me if you answer this question in reference to just one or two concrete classification tasks—whichever you find convenient. For instance, if you pick up OCR (optical character recognition, e.g., as explained in Michael Nielson’s free online book [^]), then you have to explain why an ANN-based OCR algorithm works in classifying those MNIST digits / alphabets.


Hint: Studies of Vedic literature won’t help. [I should know!] OTOH, studies of good books on epistemology, or even just good accounts covering methods of science, should certainly come in handy.

I will give you all some time before I come back on that question.

In the meanwhile, have fun—if you wish to, and of course, if you are able to. With questions of this kind. (Translating the emphasis in the italics into chaste Marathi: “laayaki asali tar.” Got it?)


A song I like:
(Marathi) “ooncha nicha kaahi neNe bhagawant”
Lyrics: Sant Tukaram
Music and Singer: Snehal Bhatkar

 

Advertisements

And to think…

Many of you must have watched the news headlines on TV this week; many might have gathered it from the ‘net.

Mumbai—and much of Maharashtra—has gone down under. Under water.

And to think that all this water is now going to go purely to waste, completely unused.

… And that, starting some time right from say February next year, we are once again going to yell desperately about water shortage, about how water-tankers have already begun plying on the “roads” near the drought-hit villages. … May be we will get generous and send not just 4-wheeler tankers but also an entire train to a drought-hit city or two…

Depressing!


OK. Here’s something less depressing. [H/t Jennifer Ouellette (@JenLucPiquant) ]:

“More than 2,000 years ago, people were able to create ice in the desert even with temperatures above freezing!” [^]

The write-up mentions a TED video by Prof. Aaswath Raman. Watched it out of idle interest, checked out his Web site, and found another TED video by him, here [^]. Raman cites statistics that blew me!

They spend “only” $24 billion on supermarket refrigeration (and other food-related cooling), but they already spend $42 billion on data-center cooling!!


But, any way, I did some further “research” and landed at a few links, like the Wiki on Yakhchal [^], on wind-catcher [^], etc.  Prof. Raman’s explanation in terms of the radiative cooling was straight-forwards, but I am not sure I understand the mechanism behind the use of a qanat [^] in Yakhchal/windcatcher cooling. It would be cool to do some CFD simulations though.


Finally, since I am once again out of a job (and out of all my saved money and in fact also into credit-card loans due to some health issue cropping up once again), I was just idly wondering about all this renewable energy business, when something struck me.


The one big downside of windmills is that the electricity they generate fluctuates too much. You can’t rely on it; the availability is neither 24X7 nor uniform. Studies in fact also show that in accommodating the more or less “random” output of windmills into the conventional grid, the price of electricity actually goes up—even if the cost of generation alone at the windmill tower may be lower. Further, battery technology has not improved to such a point that you could store the randomly generated electricity economically.

So, I thought, why not use that randomly fluctuating windmill electricity in just producing the hydrogen gas?

No, I didn’t let out a Eureka. Instead, I let out a Google search. After all, the hydrogen gas could be used in fuel-cells, right? Would the cost of packaging and transportation of hydrogen gas be too much? … A little searching later, I landed at this link: [^]. Ummm… No, no, no…. Why shoot it into the natural gas grid? Why not compress it into cylinders and transport by trains? How does the cost economics work out in that case? Any idea?


Addendum on the same day, but after about a couple of hours:

Yes, I did run into this link: “Hydrogen: Hope or Hype?” [^] (with all the links therein, and then, also this: [^]).

But before running into those links, even as my googling on “hydrogen fuel energy density” still was in progress, I thought of this idea…

Why at all transport the hydrogen fuel from the windmill farm site to elsewhere? Why not simply install a fuel cell electricity generator right at the windmill farm? That is to say, why not use the hydrogen fuel generated via electrolysis as a flywheel of sorts? Get the idea? You introduce a couple of steps in between the windmill’s electricity and the conventional grid. But you also take out the fluctuations, the bad score on the 24X7 availability. And, you don’t have to worry about the transportation costs either.

What do you think?


Addendum on 12th July 2018, 13:27 hrs IST

Further, I also browsed a few links that explore another,  solution: using compressed air: a press report [^], and a technical paper [^]. (PDF of the paper is available, but the paper would be accessible only to mechanical engineers though. Later Update: As to the press report, well, the company it talks about has already merged with another company, and has abandoned the above-ground storage of compressed air [^])

I think that such a design reduces the number of steps of energy conversions. However, that does not necessarily mean that the solution involving hydrogen fuel generation and utilization (both right at the wind-farm) isn’t going to be economical.

Economics determines (or at least must determine) the choice. Enough on this topic for now. Wish I had a student working with me; I could have then written a paper after studying the solution I have proposed above. (The idea is worth a patent too. Too bad I don’t have the money to file one. Depressing, once again!!)


OK. Enough for the time being. I may later on add the songs section if I feel like it. And, iterative modifications will always be done, but will be mostly limited to small editorial changes. Bye for now.