Hi! I’m Ameya, and since you’re reading this: a-may-yeah (as in a book, may I?, yeah sure!). “Ameya” is the sanskrit word “अमेय” approximately translating to “Impossible”. Just too full of energy as a baby, right? My parents went from “You’re impossible” to “You’re Ameya”. This github username was created when in high school, as a battle against the impossible dream of being “Dr. Ameya” (pun intended) and motivation for it.
I love to think about technology and mechanism design– How to better bake nicer incentives into decentralized technologies for a fairer world. I aspire to an the ideal of “theoretically grounded approaches should go hand-in-hand with challenging practical problems” in my research: Principled approaches can accelerate research compared to extensive hit-and-try approaches popular today. Conversely, it’s important to make real-world assumptions to model computationally hard problems and side-step worst-case complexity. As Marx once famously said “Those are my principles, and if you don’t like them.. well, I have others”. I work on effective methods for addressing continual learning at Torr Vision Group, University of Oxford.
On the fun side, I love fake frameworks: afterall ontologies make things seem real. The jungian personality type of the academic me is currently ENTJ, dominated by Te and Ni; or my MTG Color Wheel results have been dominated by Blue and Red with a Black tinge. I do read a lot of aspiring-rationalist, effective-altruism, pop-economics blogrolls. I love EDM music and everything around it: I am a huge fan of house, techno & trance music, I play a launchpad (Novation MK2) and am learning to shuffle. I have weird political leaning that I think more people should have: I am an anarchist in interpersonal interactions, communist at home, socialist democrat at local/regional level and classical liberal at the national/global level. The issues I deeply care about are animal welfare, digital privacy, slow science or the free and open source culture. If this last paragraph interests you more than anything above it, here is the non-academic me where you might find more of the same. Welcome to my personal page folks!
| ECCV20 |
Deep Expander Networks
Hybrid Binary Networks
Hybrid Binary Networks
- Shyam, the first master’s student I worked with is pursuing a PhD! Congratulations Shyam!
- Recent work with Shyam on simple, scalable ways of incorporating heirarchies to incorporate a sense of ‘degree of badness’ of mistakes in deep networks No Cost Likelihood Manipulation at Test Time for Making Better Mistakes in Deep Networks got accepted in ICLR 2021!
- Recent work with Aurobindo on effectively merging pruning and binarization paradigms STQ-Nets: Unifying Network Binarization and Structured Pruning got accepted in BMVC 2020!
- Served as reviewer for ECCV 2020. Got selected for the oustanding reviewer award!
- My recent work on problems with continual learning GDumb: A Simple Approach that Questions Our Progress in Continual Learning got accepted at ECCV 2020 (Oral)
- Served as reviewer for CVPR 2020. Got selected for the outstanding reviewer award!
- I’m pursuing a PhD focused on Machine Learning at University of Oxford, one small step towards achieving my big goal!
- My recent work on problems with active learning Sampling Bias in Deep Active Classification: An Empirical Study got accepted at EMNLP 2019
- Served as a reviewer for ICCV 2019 and AAAI 2019
- Completed my bachelors and masters in computer science at IIIT-Hyderabad. My thesis focused on developing principled approaches for sparse connectivity (pruning) and methods to exploit fast-binary convolutions for building efficient but accurate deep networks
- Starting reviewing independently. Served as a reviewer for CVPR 2019. Got outstanding reviewer award for CVPR 2019! Here is the reviewer guideline I developed for reference
- Got a yearlong residency at Verisk HQ, to work for the ML group of a NASDAQ 100 financial risk assessment firm
- I attended ECCV 2018. Thank you Google for providing full travel-grant!
- My research on Pruning Deep Expander Networks: Efficient Networks from Graph Theory got accepted at ECCV 2018 (Oral)
- I completed a research internship at IBM Research to experience industrial research. Got Best Intern Award, Summer ‘18 for outstanding research at IBM
- I attended WACV 2018 at Lake Tahoe, USA. Thank you CVIT for providing full-funding!
- My research on Hybrid Binary Networks: Optimizing for Accuracy, Efficiency and Memory got accepted at WACV 2018
- My research on Distribution-Aware Binarization of Neural Networks for Sketch Recognition got accepted at WACV 2018
- Got selected for Dean’s Merit List for Academic Excellence Award- 2018
- Got Undergraduate Research Award for Outstanding Publications in Undergraduate years -2018
- Got selected for Dean’s Merit List for Academic Excellence Award- 2017
- Got Undergraduate Research Award for Outstanding Publications in Undergraduate years -2017