🚨New Dataset Distillation paper 🚨
We propose PoDD, a new dataset distillation setting for a tiny🤏under 1 image-per-class (IPC) budget❗In this CIFAR100 example, current SoTA is 35.5% acc using ~100k pixels, PoDD gets 35.7% with ~40k pixels
Project: vision.huji.ac.il/podd
🧵👇

Our final lecture on diffusion models:
youtube.com/watch?v=bWPzUx…

We cover how to condition an image generation using text or other information (e.g. image sketch) as well as how to speed up training/inference.

Next stop: LLMs ... Stay tuned!

How to choose a vision model for your specific needs?

How do ConvNet / ViT, supervised / CLIP models compare with each other on metrics beyond ImageNet?

Our work comprehensively compares common vision models on 'non-standard' metrics. (1/n)

Classic generative vs discriminative debate has been long-standing in the ML community.

Our #NeurIPS2023 paper shows how to get the best of both worlds -- use the generative model feedback to adapt a discriminative model at test time. Robust on OOD data!

diffusion-tta.github.io

New YouTube video: 1hr general-audience introduction to Large Language Models
youtube.com/watch?v=zjkBMF…

Based on a 30min talk I gave recently; It tries to be non-technical intro, covers mental models for LLM inference, training, finetuning, the emerging LLM OS and LLM Security.

Today, we are releasing Stable Video Diffusion, our first foundation model for generative AI video based on the image model, Stable Foundation. As part of this research preview, the code, weights, and research paper are now available.

Additionally, today you can sign up for our…

You probably have a vague idea of how diffusion models work: progressively add noise to images during training, and learn a model to remove that noise. To generate a new image, just denoise a 'pure noise' image. Let's provide some intuition to the formulas involved. 🔽

Greg Brockman

Turn any paper into audio and listen on your phone!

Listen while walking, making dinner, or exercising. It's easy to fit your learning into your daily routine!

EfficientML.ai Lecture 16 - Diffusion Models (MIT 6.5940, Fall 2023)
Denoising diffusion models
Conditional diffusion
Latent diffusion
Image editing
Model personalization
Fast sampling
DDIM
Progressive distillation
Guided distillation

youtu.be/nFE1euQ_Wtw

New on arXiv!!! 🚨🚨🔥🔥🤯🤯

A Unified Framework to Enforce, Discover, and Promote Symmetry in Machine Learning
by Sam Otto, with N. Zolman, N. Kutz, & myself

A tour de force in machine learning & differential geometry

arxiv.org/abs/2311.00212

(w/ beautiful drawings by Sam)

How to write math in a paper?

Math allows you to convey your idea precisely and concisely. But how to write them clearly? 🤔

Check out some high-level tips (with examples). 🧵

Enjoyed visiting UC Berkeley’s Machine Learning Club yesterday, where I gave a talk on doing AI research. Slides: docs.google.com/presentation/d…

In the past few years I’ve worked with and observed some extremely talented researchers, and these are the trends I’ve noticed:

1. When…

🌟[NEW blog post] - Best #ICCV paper highlights: top papers summarized here: theaisummer.com/iccv-2023/

#ICCV2023 #ICCV 2023 #deeplearning #MachineLearning ICCV2021 Google AI AI at Meta NVIDIA AI AI Summer Sergios Karagiannakos

Generalized Sum Pooling for Metric Learning
arxiv.org/abs/2308.09228
ICCV 2023

A learnable generalized sum pooling method that enhances global average pooling by selectively focusing on relevant semantic entities and weighing their importance.

Are you (or is someone you know) teaching AI / ML / Deep Learning this year? My forthcoming book (freely available at udlbook.com) will save you a lot of time. This thread will show you why.

How can we train generative models on 100 examples without pretraining on large datasets? Presenting Adaptive IMLE, which outperforms both GANs and diffusion models. Details at mehranagh20.github.io/AdaIMLE/.

Work w/ Mehran Aghabozorgi and Shichong Peng. Come by our #ICML2023 poster Wed AM

1/2

Excited to turn all the memes into an actual course UMD Department of Computer Science.

Looking forward to developing this course! 🤩

This is huge: Llama-v2 is open source, with a license that authorizes commercial use!

This is going to change the landscape of the LLM market.
Llama-v2 is available on Microsoft Azure and will be available on AWS, Hugging Face and other providers

Pretrained and fine-tuned…

Have you ever wondered if generative models like #diffusion models inherently can solve discriminative tasks?

Presenting our work:
'Diffusion Models Beat GANs on Image Classification'

arxiv: arxiv.org/abs/2307.08702
website: mgwillia.github.io/diff-ssl
hf: huggingface.co/papers/2307.08…🧵