Mélanie Ducoffe retweeted
Open Problems in Applied Deep Learning
If you're looking for interesting open problems in DL, this is a good reference.
Not sure if intentional but it also looks useful to get a general picture of current trends in deep learning with ~300 references.
arxiv.org/abs/2301.11316
Mélanie Ducoffe retweeted
Introducing Make-A-Video3D! Generating 3D dynamic (mini) scenes from input text.
That is, text --> 4D!
Needs no 4D data (i.e., no dynamic 3D data), no static 3D data, no paired text-video data.
Paper: arxiv.org/abs/2301.11280
Website: make-a-video3d.github.io/
Mélanie Ducoffe retweeted
Are you a young researcher in mathematics or computer science? Take part in the 10th HLF and get the chance to interact and network with other young researchers from all over the world and laureates of the most prestigious awards.
Apply now: application.heidelberg-laure…
Mélanie Ducoffe retweeted
Scikit-Learn and 🤗 join forces!
With a growing number of tabular classification & regression checkpoints, we believe statistical ML has its place on the HF Hub.
We're excited to partner with sklearn, statistical ML champion, and move forward together.
blog.scikit-learn.org/update…
Mélanie Ducoffe retweeted
Gather round, Twitter folks, it's time for our beloved
**Alice's adventures in a differentiable wonderland**, our magical tour of autodiff and backpropagation. 🔥
Slides below 1/n 👇
Mélanie Ducoffe retweeted
Today, along with my collaborators at @GoogleAI, we announce DreamBooth! It allows a user to generate a subject of choice (pet, object, etc.) in myriad contexts and with text-guided semantic variations! The options are endless. (Thread 👇)
webpage: dreambooth.github.io/
1/N
Mélanie Ducoffe retweeted
We (@BeEngelhardt, @NailaMurray and I) are proud to announce the creation of a Journal-to-Conference track, in collaboration with JMLR and conferences NeurIPS 2022, ICLR 2023 and ICML 2023!
neurips.cc/public/JournalToC…
iclr.cc/public/JournalToConf…
icml.cc/public/JournalToConf…
Mélanie Ducoffe retweeted
If you are at ICML 2022 #ICML2022 , please stop by our workshop on Formal Verification of Machine Learning (WFVML) today (Friday) in Room 308! See workshop schedule at ml-verification.com/schedule
Could not agree more teapawt.org/mvechev/status/1504063…
Replying to @ylecun
I think combining ideas from differentiable and symbolic is the way to go :) Many core ideas from symbolic (abstraction, modularity, fixed point reasoning, learning with small data, etc.) are super interesting to adapt and fuse with DL and are used in industry at scale.
Awesome introduction to BERT. teapawt.org/BritneyMuller/status/1…
🎯 Goal: Make BERT as accessible as possible
✏️ Content: BERT 101 - State of the Art NLP Model Explained: huggingface.co/blog/bert-101
📔 Colab Notebook: Getting Started with BERT (included)
1
Mélanie Ducoffe retweeted
Don't Lie to Me! Robust and Efficient Explainability with Verified Perturbation Analysis
abs: arxiv.org/abs/2202.07728
Mélanie Ducoffe retweeted
Bon g fait ma petite investigation en appelant ttes ls agences immobilières de ma ville notamment celle qui a vendu le bien quelques étages au-dessus de chez nous. Nous étions, au début, sur un prix complètement déconnant au moment de contractualiser soit une perte de 2/3semaines
Mélanie Ducoffe retweeted
This lecture by @MihaelaCRosca & Jeff Donahue looks at the theory behind GANs, the difficulties involved in optimising them & theoretical and empirical improvements to the basic framework.
https://teapawt.org/t.co/UG3BJSjzld #AtHomeWithAI
Mélanie Ducoffe retweeted
New work on creating differentiable versions of optimizers, with perturbation techniques. Can be used to differentiate through solutions of opt. problems, and to learn from examples and their solutions in many tasks (e.g. classif, ranking, shortest paths) arxiv.org/abs/2002.08676
Mélanie Ducoffe retweeted
If you would like to recommend or become a reviewer for NeurIPS 2020, please fill out the form linked below. Thanks!
docs.google.com/forms/d/11N0…
Mélanie Ducoffe retweeted
My slides: drive.google.com/file/d/1r-m…
Video of the talks: vimeo.com/390347111
- 1:10 in for Geoff Hinton's keynote,
- 1:44 for mine,
- 2:18 for Yoshua Bengio's,
- 2:51 for the panel discussion moderated by Leslie Pack Kaelbling
Mélanie Ducoffe retweeted
Classic numerical methods like Jacobi, Block-Jacobi, and its hybrid with GaussSeidel are useful for accelerating ML. Using them for sampling makes your trained autoregressive models run much faster "for free"! Interested in more such applications? Please read our latest work.



























