You use GPUs everyday, but do you (actually) know how they work?
— Sasha Rush (@srush_nlp) July 12, 2022
GPU-Puzzles (v0.1) - 14 short puzzles in Python with a visual debugger. No background required. Do puzzles, learn CUDA.
Link: https://t.co/Yk1lWRqilN pic.twitter.com/eFs7u5lxES
A check-in at the 1-3 year mark before this prediction expires. With
— Sebastian Raschka (@rasbt) July 9, 2022
(1) data-centric AI building successful ML systems across health care, government tech, and manufacturing;
(2) large language models enabling breakthroughs in language modeling,
are we still on track with this? https://t.co/0NXhCCl9kq
"torch.manual_seed(3407) is all you need: On the influence of random seeds in deep learning architectures for computer vision" https://t.co/vP0RuImY8e haha. Actually torch.cuda.manual_seed is also what you need. But clearly 3407 looks like the top rng seed to use :)
— Andrej Karpathy (@karpathy) July 8, 2022
Spoke to a data scientist who was frustrated that his company decided to buy a proprietary demand forecasting system when the performance of his XGBoost model was so much stronger.
— Anthony Goldbloom (@antgoldbloom) July 8, 2022
His solution: he fed his XGBoost forecasts into the proprietary system as a feature 😂
What a chart! This chart visualizes the frequency of compound insults (e.g. poophead, scumwad) in Reddit comments in the shape of a prefix and suffix matrix. This is super helpful in creating new swearwords too. Perfect for the social media age. Source: https://t.co/ghKHfGBtWP pic.twitter.com/vLwZh5QY1Y
— Simon Kuestenmacher (@simongerman600) July 7, 2022
At @NVIDIA we have developed Merlin, an end-to-end Recommender System toolbox that can simplify and accelerate RecSys deployment. Read more about it in this thoughtful and informative piece by @radekosmulski and Benedikt Schiffererhttps://t.co/YYt2duWFXD#RecSys #ds #ml #ai 3/3
— Bojan Tunguz (@tunguz) July 7, 2022
New Typer release! 0.5.0 🔖
— Sebastián Ramírez (@tiangolo) July 6, 2022
This one has pretty output when the Typer app (your code) has errors. ✨
Now @textualizeio's (@willmcgugan) Rich is an optional dependency, used automatically by Typer. 🎉
Install with:
pip install "typer[all]"
More examples below 👇 pic.twitter.com/7UQbxMFtbb
MLGO, a general industrial-grade #ML framework for compiler optimization, uses #RL to train ML policies to replace complicated heuristics, reducing code size and improving performance. Learn more and review the open-source training solution ↓ https://t.co/ztRsRkr3Ka
— Google AI (@GoogleAI) July 6, 2022
Neural Networks and the Chomsky Hierarchy
— AK (@_akhaliq) July 6, 2022
abs: https://t.co/u6Jl2WvKMr
sota architectures, such as LSTMs and Transformers, cannot solve seemingly simple tasks, such as duplicating a string, when evaluated on sequences that are significantly longer than those seen during training pic.twitter.com/Y3SCESehTN
2️⃣ Hopular (Schäfl et al) - proposes a deep learning architecture based on continuous Hopfield networks for competitive results on small-sized tabular datasets.https://t.co/mAuZYyryXb pic.twitter.com/nlctJ5ZT6j
— Papers with Code (@paperswithcode) July 5, 2022
"Pen & Paper Exercises in Machine Learning" -- probably the most interesting arXiv article I've seen in a while ☺️
— Sebastian Raschka (@rasbt) July 1, 2022
📝 https://t.co/JQWhJadxzo
Love it when exercises come with solutions. 👌
(Textbooks with exercises but w/o solutions are truly evil 😆). pic.twitter.com/bULUF1UclK
Forecasting Future World Events with Neural Networks
— AK (@_akhaliq) July 1, 2022
abs: https://t.co/tD8F0ZC1rC
github: https://t.co/v8HZgye0ZH
a dataset for measuring the ability of neural networks to forecast future world events, containing thousands of forecasting questions and an accompanying news corpus pic.twitter.com/xsQnxfgdia
