Deep Learning is a form of human-assisted but mostly constraint-driven software development. It works because a particular smooth relaxation of program space allows a surprisingly efficient and effective local search. Something like that, my favorite definition.
— Andrej Karpathy (@karpathy) September 19, 2021
We're preparing a short course for PhD students on machine learning security, and open sourcing the content. Any feedback is more than welcome -- towards improving next year's extended edition!https://t.co/aXbkRzaR0V
— Battista Biggio (@biggiobattista) September 18, 2021
Googling Tip: when you want to learn some ML concept, google "python <concept name> from scratch".
— Andrew Trask (@iamtrask) September 17, 2021
This will often lead to a blogpost with:
- intuitive explanation
- a toy implementation
- reference to more formal papers
which you can unpack in that order :)#100DaysOfMLCode
A very sobering map by @atlasmapper shows the total number of deaths suffered by European countries during WWII. About 3% of the global population was killed during WWII. Source: https://t.co/b0YvNFeevG pic.twitter.com/4T7ytwbh54
— Simon Kuestenmacher (@simongerman600) September 17, 2021
Wish your neural networks faster and more accurate?
— Mingxing Tan (@tanmingxing) September 16, 2021
Check out our recent EfficientNetV2 and CoAtNet, which significantly speed up the training and inference, while achieving state-of-the-art 90.88% top-1 accuracy on ImageNet. https://t.co/9buCSZmYby
What is your favorite tool for labeling data? Labelme (for image data) came to mind, but then going down the rabbit hole of this question, I learned that there is an entire "awesome-" GitHub repo of data labeling tools: https://t.co/w7ZApH9hT1
— Sebastian Raschka (@rasbt) September 16, 2021
For downloading large image datasets (1M+), I highly recommend https://t.co/U27VlPBfUK from @rom1504
— Boris Dayma 🥑 (@borisdayma) September 15, 2021
You can even monitor performance and download errors with @weights_biases pic.twitter.com/ZvrHh6B8O0
📢 Introducing 🤗 Optimum
— Hugging Face (@huggingface) September 14, 2021
A new open source library to optimize 🤗Transformers for production performance. 🏎
Quantize, Prune, Optimize models easily, targeting hardware from our partners @intel @graphcoreai @Qualcomm! 🤩https://t.co/oemVDWlnxI
An Empirical Study of GPT-3 for Few-Shot Knowledge-Based VQA
— AK (@ak92501) September 13, 2021
abs: https://t.co/A3s6zdW0Iu
a simple yet effective method that Prompts GPT3 via the use of Image Captions. Using only 16 examples, PICa surpasses the supervised sota by an absolute +8.6 points on the OK-VQA dataset pic.twitter.com/9HKybsk1qu
A better Dropout! Implementing DropBlock in PyTorchhttps://t.co/OXUUq31rtM #deeplearning #machinelearning #ml #ai #neuralnetworks #datascience #PyTorch pic.twitter.com/mhS7Hnq5yy
— PyTorch Best Practices (@PyTorchPractice) September 12, 2021
LAION-400M: open-source dataset of 400 million image-text pairs
— AK (@ak92501) September 12, 2021
project page: https://t.co/IA8aNpXZ6a pic.twitter.com/f5IoLESnRx
For all you *aspiring* @PyTorch users!@kaixhin has written a *very* nice fast-track intro!
— Andrew Trask (@iamtrask) September 12, 2021
(and it appears he's keeping it up to date!)#100DaysOfMLCode #100DaysOfCodehttps://t.co/bEILhwVkib
