KL divergence asymmetry

Grateful to have our work accepted in ECCV 2020 with amazing co-authors Tian Han, Bo Pang, et al., "Learning multi-layer latent variable model via variational optimization of short run MCMC for approximate inference": arxiv.org/pdf/1912.01909… #ECCV2020 #mcmc

NVIDIA DRIVE Sim Software Built on NVIDIA Omniverse youtu.be/DuK4ppn0g3A via YouTube

I’m having an especially hard time concentrating on research this week, but instead of doomscrolling on twitter I guess I’ll do something semi-productive and make another twitter thread. Today's topic: What is a Lyapunov function? 1/

Late announcement: A review paper "Representation Learning: A Statistical Perspective" arxiv.org/abs/1911.11374 was published by Annual Reviews with Ruiqi Gao Erik Nijkamp.The paper covers the current hot topics, e.g. ConvNet-EBM, amortized sampling for EBM, grid-cell model, etc.

Really enjoyed Szymon Migacz's talk on optimizing PyTorch's comp. performance deep learning applications: youtube.com/watch?v=9mS1fI… Summarizing the main points: (1) parallel data processing via num_workers > 0 (2) better memory GPU utilization via pin_memory = True 1/4

Check out our work 3D Shape Generation and Completion through Point-Voxel Diffusion (PVD)! PVD probabilistically and diversely completes partial shapes from (real) depth scans and achieves SOTA on shape generation. w/ Yilun Du and Jiajun Wu Website: alexzhou907.github.io/pvd

Finally! The slides of the CVPR 2021 tutorial on energy-based models #EBM are available now: energy-based-models.github.io/document/EBMtu…

Our model CodeGen (up to 16.1B trained on TPU-v4) is open-source. Competitive with OpenAI Codex. Paper: arxiv.org/abs/2203.13474 Code: github.com/salesforce/Cod… Wonderful team: Bo Pang, Hiroaki_Hayashi, Lifu Tu, Huan Wang , Yingbo , Silvio Savarese, Caiming Xiong.

The best lesson on writing structure I've ever read:

Introducing: AI Powered 3D Search 🔍 We've created the fastest search on the most expansive library of 3D assets With almost 1 million searchable assets and an average search time of 50ms, it is easier than ever to find the perfect 3D asset for your project Try out the magic✨

I marvel at MidJourney's ability to produce art of historical battles. It's time for another AI thread you never asked for. Famous Historical Battles The Fall of Constantinople

New openings NVIDIA for Genomic LLM R&D at sr levels. You will build genomic foundation models into our products and see your work used by the most iconic techbios. Learn more about BioNemo: m.youtube.com/watch?v=GNL1z7… Apply here: nvidia.wd5.myworkdayjobs.com/en-US/NVIDIAEx…

This is extremely well said. From NYC Mayor Adam’s

I unsubscribed from Dropbox many years ago and they have kept my data. They store it for free and when I want it back I can re-subscribe.

📢🔥 My team at NVIDIA Research is looking for full-time research scientists & summer interns. Topics of interest are: 1⃣Gen AI for science (climate, biology, chemistry) 2⃣Image/Video/3D (generate, edit, manipulate) 3⃣Fundamental generative learning Apply via links below

The best recommendation systems in use today aim to capture your attention and waste your time, maximizing profits for advertising companies like Google, Meta, or TikTok. Imagine instead a powerful algorithm on your side. A state-of-the-art system that learns from your behavior

Whoever figures out how to create high protein, pre made pizza dough and sell it in the freezer at grocery stores is going to be a millionaire. This entire pizza is 700 calories and is 64g of protein, 74g of carbs, and 15g of fat.