Jyotirmoy Bhattacharya
@yomritoyj.bsky.social
Student and teacher of economics. Math, computing and physics enthusiast. Associate Professor, Ambedkar University Delhi. https://jyotirmoy.net
As someone interested in formalism, I don't understand why logic textbook authors are so defensive about it. Like, “no, no, no, we are not the syntactic Taliban, we believe in mathematical truth.” Even when it is not necessary for what they are doing.
October 27, 2025 at 4:56 AM
As someone interested in formalism, I don't understand why logic textbook authors are so defensive about it. Like, “no, no, no, we are not the syntactic Taliban, we believe in mathematical truth.” Even when it is not necessary for what they are doing.
Fitness sites: Don't check your weight everyday.
Me: Without sufficient recent data on the noise process how can I ensure the optimality of my filter?
{I'm on the Terence Tao diet: An epsilon of calorie deficit. So efficient signal extraction is important.}
Me: Without sufficient recent data on the noise process how can I ensure the optimality of my filter?
{I'm on the Terence Tao diet: An epsilon of calorie deficit. So efficient signal extraction is important.}
September 23, 2025 at 8:09 AM
Fitness sites: Don't check your weight everyday.
Me: Without sufficient recent data on the noise process how can I ensure the optimality of my filter?
{I'm on the Terence Tao diet: An epsilon of calorie deficit. So efficient signal extraction is important.}
Me: Without sufficient recent data on the noise process how can I ensure the optimality of my filter?
{I'm on the Terence Tao diet: An epsilon of calorie deficit. So efficient signal extraction is important.}
Since the common refinement of two axis-aligned partitions is itself an axis-aligned partition, any predictor based on a finite ensemble of decision trees can be replaced by a single decision tree.
February 22, 2025 at 5:41 AM
Since the common refinement of two axis-aligned partitions is itself an axis-aligned partition, any predictor based on a finite ensemble of decision trees can be replaced by a single decision tree.
I thought I was competent to teach Python. Then I was asked to explain this.
February 12, 2025 at 9:20 AM
I thought I was competent to teach Python. Then I was asked to explain this.
If A is 1×10, B is 10×50 and C is 50×1000, then (AB)C is much cheapter to calculate than A(BC).
Used this to argue why neural networks use reverse mode autodiff rather than forward mode (the matrices are the Jacobians of the layers).
Used this to argue why neural networks use reverse mode autodiff rather than forward mode (the matrices are the Jacobians of the layers).
February 6, 2025 at 12:55 PM
If A is 1×10, B is 10×50 and C is 50×1000, then (AB)C is much cheapter to calculate than A(BC).
Used this to argue why neural networks use reverse mode autodiff rather than forward mode (the matrices are the Jacobians of the layers).
Used this to argue why neural networks use reverse mode autodiff rather than forward mode (the matrices are the Jacobians of the layers).
Had a class on numpy and gave a lecture instead on clock frequency*ipc*simd and memory hierarchy. Who cares what it is calculating as long as you can make it go vroom :)
January 22, 2025 at 3:31 PM
Had a class on numpy and gave a lecture instead on clock frequency*ipc*simd and memory hierarchy. Who cares what it is calculating as long as you can make it go vroom :)
I want to pack all four months' data for a household from a CPHS wave into a single vector just so that I can call it a wave packet.
January 15, 2025 at 3:52 PM
I want to pack all four months' data for a household from a CPHS wave into a single vector just so that I can call it a wave packet.
First draft of my lecture notes chapter what I find to be to the most difficult part of ML. The problem of induction. Would greatly appreciate comments. www.jyotirmoy.net/posts/2024-1...
ML Notes 2025. Supervised Learning and the Problem of Induction
www.jyotirmoy.net
December 30, 2024 at 7:51 AM
First draft of my lecture notes chapter what I find to be to the most difficult part of ML. The problem of induction. Would greatly appreciate comments. www.jyotirmoy.net/posts/2024-1...
Thanks to a work project, I discovered how expensive it is to control a touch monitor from more than 2m away. I'd have set up a remote desktop over WiFi and called it a day, but apparently it is considered too fragile by the pros. www.avaccess.com/products/4ke...
AV Access 4K KVM Extender w/ HDMI 1.4 & Zero Latency
AV Access 4KEX100-KVM is a 4K KVM extender over HDBaseT. It can transmit 4K@60Hz HDMI signal over a single Cat6 cable up to 330ft/100m with zero loss or delay. It also supports 1080P@120Hz, ideal for ...
www.avaccess.com
December 10, 2024 at 4:42 PM
Thanks to a work project, I discovered how expensive it is to control a touch monitor from more than 2m away. I'd have set up a remote desktop over WiFi and called it a day, but apparently it is considered too fragile by the pros. www.avaccess.com/products/4ke...
They are Eliza-fying Claude. Seeing a sharp uptick in the number of responses that end with a question to me. Must everything succumb at the altar of Engagement?
November 28, 2024 at 6:19 AM
They are Eliza-fying Claude. Seeing a sharp uptick in the number of responses that end with a question to me. Must everything succumb at the altar of Engagement?
Got the new translation. At first sight the modernization of the language reads well, though I hope they have not smoothened away all of the old man's fire and brimstone. Paper too thin, print shows through @princetonupress.bsky.social
November 26, 2024 at 2:32 PM
Got the new translation. At first sight the modernization of the language reads well, though I hope they have not smoothened away all of the old man's fire and brimstone. Paper too thin, print shows through @princetonupress.bsky.social
Wouldn't it be better to rename exp() to e()?
November 24, 2024 at 5:55 AM
Wouldn't it be better to rename exp() to e()?
Reposted by Jyotirmoy Bhattacharya
For the second time in my career, someone has tried to seriously insult me by saying that my research is more sociology than proper history.
I could never find that insulting.
Quite the opposite, actually.
I could never find that insulting.
Quite the opposite, actually.
November 20, 2024 at 8:48 AM
For the second time in my career, someone has tried to seriously insult me by saying that my research is more sociology than proper history.
I could never find that insulting.
Quite the opposite, actually.
I could never find that insulting.
Quite the opposite, actually.
Experimenting in mixing Greek and Devanagari letters in the teaching of macro
November 20, 2024 at 8:11 AM
Experimenting in mixing Greek and Devanagari letters in the teaching of macro
Claude (new Sonnet) is fun to chat with, but it let me down again yesterday when I tried to do serious work with it, making errors in solving quadratic and linear equations, producing LaTeX that did not compile. Also the very smallness of the output size limit becomes a pain too quickly.
November 17, 2024 at 6:30 AM
Claude (new Sonnet) is fun to chat with, but it let me down again yesterday when I tried to do serious work with it, making errors in solving quadratic and linear equations, producing LaTeX that did not compile. Also the very smallness of the output size limit becomes a pain too quickly.
Reposted by Jyotirmoy Bhattacharya
there are *surprisingly many* open problems when it comes to theory/methods in causal inference
check out this talk by Siva Balakrishnan for an excellent & comprehensive summary of the state of the art
www.youtube.com/live/Mnum0Ox...
www.stat.cmu.edu/~siva/
check out this talk by Siva Balakrishnan for an excellent & comprehensive summary of the state of the art
www.youtube.com/live/Mnum0Ox...
www.stat.cmu.edu/~siva/
November 12, 2024 at 3:15 AM
there are *surprisingly many* open problems when it comes to theory/methods in causal inference
check out this talk by Siva Balakrishnan for an excellent & comprehensive summary of the state of the art
www.youtube.com/live/Mnum0Ox...
www.stat.cmu.edu/~siva/
check out this talk by Siva Balakrishnan for an excellent & comprehensive summary of the state of the art
www.youtube.com/live/Mnum0Ox...
www.stat.cmu.edu/~siva/
I'm only now reading the PoCo-PoMo literature of the late 20th century — Partho Chatterjee, Ashish Nandy, Dipesh Chakrabarty etc — and finding it a hoot.
It seems to be a amalgam of the two fundamental Indian responses to Western superiority. (1/n)
It seems to be a amalgam of the two fundamental Indian responses to Western superiority. (1/n)
November 16, 2024 at 8:03 AM
I'm only now reading the PoCo-PoMo literature of the late 20th century — Partho Chatterjee, Ashish Nandy, Dipesh Chakrabarty etc — and finding it a hoot.
It seems to be a amalgam of the two fundamental Indian responses to Western superiority. (1/n)
It seems to be a amalgam of the two fundamental Indian responses to Western superiority. (1/n)
Reposted by Jyotirmoy Bhattacharya
I did a thread a year or so ago, on the bad site, about how marigolds were a late insertion into Hindu rituals (late = 15th-16th centuries).
Hindu nationalists hounded me for weeks over it.
www.nytimes.com/2024/11/14/t... #Hinduism #Hindutva #ritual
Hindu nationalists hounded me for weeks over it.
www.nytimes.com/2024/11/14/t... #Hinduism #Hindutva #ritual
How Marigolds Became Central to Hindu Rituals in India
Originating in North America, the blossoms have become central to Hindu rituals in India.
www.nytimes.com
November 15, 2024 at 8:16 PM
I did a thread a year or so ago, on the bad site, about how marigolds were a late insertion into Hindu rituals (late = 15th-16th centuries).
Hindu nationalists hounded me for weeks over it.
www.nytimes.com/2024/11/14/t... #Hinduism #Hindutva #ritual
Hindu nationalists hounded me for weeks over it.
www.nytimes.com/2024/11/14/t... #Hinduism #Hindutva #ritual
CTRL (2024) is just sanitised Demonlover (2002).
November 15, 2024 at 7:05 AM
CTRL (2024) is just sanitised Demonlover (2002).
Why aren't more people talking about the replication crisis in particle physics? How can the CERN people call themselves scientists when no one else in the whole world has even replicated their apparatus, to speak nothing of their results.
November 15, 2024 at 7:04 AM
Why aren't more people talking about the replication crisis in particle physics? How can the CERN people call themselves scientists when no one else in the whole world has even replicated their apparatus, to speak nothing of their results.
Give me a criteria for selecting subjects to be taught at the university which includes math but excludes astrology. Don't say usefulness, you'll offend many mathematicians.
November 15, 2024 at 2:55 AM
Give me a criteria for selecting subjects to be taught at the university which includes math but excludes astrology. Don't say usefulness, you'll offend many mathematicians.
Reposted by Jyotirmoy Bhattacharya
People from cultures where donuts are traditionally eaten: are you supposed to dunk them in coffee? I know there is a whole fast-food chain named after this, but it still does not seem like a good idea.
November 13, 2024 at 9:56 AM
People from cultures where donuts are traditionally eaten: are you supposed to dunk them in coffee? I know there is a whole fast-food chain named after this, but it still does not seem like a good idea.
AJR getting the Nobel got me into a fight with friends about capitalism and colonialism. Now Wainwright and Vershynin are gathering dust and I'm trying to figure out 'What is History?'
They should have given the prize to Athey.
They should have given the prize to Athey.
November 13, 2024 at 8:17 AM
AJR getting the Nobel got me into a fight with friends about capitalism and colonialism. Now Wainwright and Vershynin are gathering dust and I'm trying to figure out 'What is History?'
They should have given the prize to Athey.
They should have given the prize to Athey.
Reposted by Jyotirmoy Bhattacharya
Interesting. I put the exact same thing on here and on Twitter. Ratio of likes to followers here is 1/200. Ratio of likes to rollowers on Twitter is 1/166, so very close. Folow me. I post about the JFE, some of my own work, complaints (and I'm usually funny when I'm pissed off), and food.
November 10, 2024 at 7:46 PM
Interesting. I put the exact same thing on here and on Twitter. Ratio of likes to followers here is 1/200. Ratio of likes to rollowers on Twitter is 1/166, so very close. Folow me. I post about the JFE, some of my own work, complaints (and I'm usually funny when I'm pissed off), and food.