Thakur Family Chair Assistant Professor @ C-MInDS, IIT Bombay
Powai, Mumbai
Joined February 2014
- Tweets217
- Following524
- Followers331
- Likes2.8K
@mybmc No action taken on this blocked & broken drainage chamber on LJ Rd despite formal complaint, follow-ups. Please take immediate action. Serious health menace for residents. Local sewage not draining into municipal gutters ⇒ clogged sewage ⇒ rats, mosquitos, cockroaches...
1
87
We now have a simple, powerful, and stable alternative to Deep Nets
A challenge in classical kernel models is choosing the 'right' kernel
By examining NN training, we identified the modification necessary to empower kernels - FEATURE LEARNING
#NeuralNetworkFreeSince2023
What is the nature of feature learning in deep networks? We propose that neural networks recover a statistic known as the average gradient outer product (AGOP).
Github: github.com/aradha/recursive_…
arXiv: arxiv.org/abs/2212.13881
Parthe Pandit retweeted
Rahul Parhi and I wrote a tutorial style article to explain modern theory of deep learning and neural function spaces via elementary signal/image processing concepts (Fourier transform, Radon transform, L1 regularization, sparsity) arxiv.org/abs/2301.09554
do @FOXSports WC commentators get paid by the word? they talk non-stop but give out 0 information!
8
This interview of Grace Wahba is a really nice read
projecteuclid.org/journals/s…
1
1
Companies promise customers data privacy.
But they are (over)valued for the customer data they own, and can potentially sell.
So they are either lying to customers or shareholders, or worse, both!!
How can we get fair valuations of data that discount for promised privacy?
#Optimization folks help me out
I have a preconditioned GD solver for
min{x} F1(x)
Now, I want to solve
min{x} F1(x) + F2(x)
and I have the prox_F2 operator
Would the same F1 preconditioner work for a proximal gradient method?
Will I always get same fixed points?
1
Like journals, can a conference like #NeurIPS2022 say "Sorry, we couldn't find reviewers, please resubmit elsewhere"?
1
1
sample rejection email
"The PC appreciates your effort to write this paper. This year we had a record-breaking number of submissions. We regret to inform you that your manuscript could not be reviewed. Good luck with a resubmission."
2
1
This CUDA been useful for us @Amirhesam_A @nmallinar
I officially verify that the following people are really good at GPU programming (and have time to procrastinate)...
1
2
Is this inequality correct for all matrices, or just a high probability property of i.i.d. random matrices?
1