Waleed Mustafa retweetledi
When I mention “Jensen’s inequality” in my class (when we teach the flaws of the averages) some people just assume it was invented by Jensen Huang lol
Mathieu@miniapeur
English
Waleed Mustafa
93 posts
Waleed Mustafa retweetledi
Microsoft AI: we take the alignment problem very seriously
Microsoft Word:
English
Waleed Mustafa retweetledi
Adam has more citations than Robbins-Monro! Wtf
Gabriel Peyré@gabrielpeyre
Oldies but goldies: H Robbins, S Monro, A Stochastic Approximation Method, 1951. Early appearance of the stochastic gradient method, which is the workhorse of many large-scale ML methods. en.wikipedia.org/wiki/Stochasti… en.wikipedia.org/wiki/Stochasti…
English
Waleed Mustafa retweetledi
Happening today at 16:30--18:30, Hall B, Poster # 82! "Efficient Integrators for Diffusion Generative Models".
Stephan Mandt@StephanMandt
How can we accelerate sampling in a trained #diffusion model? Learn about the theoretical justifications and generalizations of DDIM and check out @kpandey008’s #ICLR2024 paper!
English
Waleed Mustafa retweetledi
Waleed Mustafa retweetledi
Amid the ceasefire, horrifying glimpses of Israeli atrocities in Gaza are slowly emerging. A 🧵to document some of the footage Israel & Biden's admin didn't want you to see:
Decomposing & charred bodies, inc. women & children piled up in the streets; entire towns decimated...
English
My son said today that “an airplane is a tricycle”. Few will appreciate how exciting this is.
English
Waleed Mustafa retweetledi
Waleed Mustafa retweetledi
@thomasahle @thegautamkamath Because we normalized characters to 20x20 pixels and centered them in the frame by the center of mass of black pixels. Because of this, the character bounding boxes could jitter by + or - 4 pixels. Hence the 28x28 frame.
English
Waleed Mustafa retweetledi
Pass by our booth. Looking forward to meeting you.
aiXplain@aixplain
We’re at #ICASSP2023. Come visit us at booth #8!
English
Waleed Mustafa retweetledi
Interesting challenge - Arabic TTS using under-resource brodcast news data. Happy to share more in @ieeeICASSP
ArabicNLP2026@_ArabicNLP
An interesting shared task for Arabic Text-to-Speech as part of #ASRU2023! Registration is currently open with the 1st of July 2023 as the tentative date for submissions 🗣️
English
Waleed Mustafa retweetledi
With all due respect, Geoff Hinton should change his Twitter bio to
"The old dude who created AI" - @SnoopDogg
Nathaniel Whittemore@nlw
Snoop breaks down AI: 1. This is what we watched as kids in movies 2. "Old guy who invented AI says its not safe cuz its got a mind of its own and these mf'ers gonna start doing their own shit" 3. Do I need to invest? 4. Does anyone know anything? Pretty damn on point.
English
Waleed Mustafa retweetledi
Thanks @aclmeeting for hosting the conference in Canada while ignoring all nationalities who won't be able to participate due to the strict visa procedure!
Given this, kindly stop claiming too much about the importance of diversity, equity & inclusion!
#ACL2023
#ACL2023NLP
English
Waleed Mustafa retweetledi
🎉Exciting news! Our "Neural Compression" workshop proposal has been accepted at #ICML 2023! Join us to explore the latest research developments, including perceptual losses and more compute-efficient models! @BerivanISIK, @YiboYang, @_dsevero, @karen_ullrich, @robamler
English
Waleed Mustafa retweetledi
In this month's blog post, Jensen's inequality, the inequality you often wish is in the other direction. francisbach.com/jensen-inequal…
English
Waleed Mustafa retweetledi
Waleed Mustafa retweetledi
Preprint is out of our AX-MABSA framework for extremely weakly supervised aspect-based sentiment analysis.
Input: unlabelled reviews + one keyword for each aspect category
Output: multiple aspect/sentiment pairs in each review
Link: arxiv.org/abs/2211.03837
#EMNLP2022
English
Waleed Mustafa retweetledi
Come stop by our poster this afternoon (1.B-101) if you want to talk with us about audio-visual ZSL! We show that our temporal and cross-modal constrained attention mechanism outperforms previous work on three audio-visual GZSL benchmarks!
#ECCV2022
Explainable Machine Learning@ExplainableML
🗓️Sess 6 | 101, 25/10 | afternoon “Temporal and cross-modal attention for audio-visual zero-shot learning” @MerceaOtniel* @hummelth_*, A. S. Koepke, @zeynepakata We leverage temporal context for better and improved audio-visual zero-shot learning! Blog bit.ly/3eWfTIg
English
Waleed Mustafa retweetledi
Patent granted 😀 🥳 🎉!
We adapted neural machine translation to fully diacritize Arabic texts with very high accuracy.
Thanks to my great colleagues: @AhmedAAbdelali, @kareem2darwish, @hassaan84s, and Younis Samih.
Paper:
aclanthology.org/N19-1248/
الحمد لله رب العالمين!
