In this thread, I want to compile a list of Deep Learning resources 📚🎬 that some people might not be aware of. It’s amazing how organizations like @fastai, @stanfordnlp and @OpenAI and people like @math_rachel ,@pieterabbeel and @lexfridman have made these freely available 🎉👇
I’ll be keeping a more comprehensive list on my GitHub. But I’m sure there are many that I missed, so I’d love to know, if you have any other suggestions! github.com/sannykim/deep-…
Most people know about the great @fastai course, but did you know @mathrachel also published a Computational Linear Algebra course with textbook and videos, covering topics such as Singular Value Decomposition and Principal Component Analysis? /1 github.com/fastai/numeric…
One course that covers more of the basic statistics for Machine Learning is @NandoDF's UBC course, which offers one of the most intuitive introductions to ML and stats /3 youtube.com/playlist?list=…
For people, who want to dive a little deeper, Gibert Strang and @MITOCW just recently published a course that combines Linear Algebra with Deep Learning and a little bit of stats and calculus /4 youtube.com/playlist?list=…
Very under-appreciated is @gneubig's CMU course, which provides another great intro NLP and Deep Learning and even covers topics such as Advanced Search Algorithms. Both courses are fantastic, so you can’t go wrong with either! /7 youtube.com/playlist?list=…
A slightly older but still invaluable resource is Oxford's and DeepMind’s 2017 Deep Learning for NLP course. Listen to @egrefen explain World Level Semantics or @redpony talk about Conditional Language Models. /8 youtube.com/playlist?list=…
Last NLP resource I'm gonna mention for now is @seb_ruder's PhD Thesis on Neural Transfer Learning for NLP, which provides an excellent overview of topics such as Multi-Task Learning and Pretrained representations in NLP. /9 ruder.io/thesis/neural_…
Jumping over to Reinforcement Learning, we have the amazing @EmmaBrunskill, who explains topics in RL in her CS234 course at Stanford. Also check out the assignments (+solutions) on the course website! /10 youtube.com/playlist?list=…
Another RL resource that's very well made is @jachiam0 and @OpenAI's "Spinning Up in RL" that contains both explanatory content and practical code examples. There's also a recorded webinar on OpenAI's YouTube channel! /11 spinningup.openai.com/en/latest/
One of the best courses recently also comes from @pabbeel's team on Deep Unsupervised Learning, which is the CS294-158 course at Berkeley. Particularly liked @AlecRad's guest lecture on Language Models (albeit the audio was sadly gone for a bit)! /14
@pabbeel has been quite amazing in contributing to online resources. So, another event he organizes with @josh_tobin_ + @sergeykarayev called Full Stack Deep Learning helps to "bridge training ML models to deploying AI systems in the real world" /15 youtube.com/playlist?list=…
Tip: There is currently a @twimlai study group on the contents of this bootcamp. Check out previous recordings and sign up for on of their regular online meetups. Props to @samcharrington and others for creating such an active community /16 youtube.com/playlist?list=…
Switching over to YouTube channels, shoutout to @ykilcher for producing some of the best paper overviews out there. He should have at least 10k subscribers! Check out his account, if you want to gain an overview of BERT, Neural ODEs etc. /20 youtube.com/playlist?list=…
When it comes to understanding papers, I love @rctatman + @kaggle's Reading Group on YouTube! She makes it really easy to get into a topic and appreciate how she addresses comments while reading these papers. Really enjoyed the videos on Transformers! /21 youtube.com/playlist?list=…
Probably the most comprehensive library of paper overviews comes from @AISC_TO (formerly TDLS), which goes through a diverse set of topics in ML. The interactive discussions, e.g. on PCGANs are super helpful to better understand a paper! /22 youtube.com/playlist?list=…
Another YouTube channel I'd like to highlight is @xsteenbrugge's Arxiv Insights! He creates very accessible introductions to concepts such as VAEs or PPO /23 youtube.com/channel/UCNIkB…
Next one on the list is @apsoleimany + @xanamini's MIT course on Intro to DL. Their Github Labs are especially nice! But even if you're familiar with these basic topics, I'd watch the guest lectures, e.g a fantastic one by @viegasf on ML visualization /24 youtube.com/playlist?list=…
One more paper reading group comes from Microsoft! Hosted by the great @ecsquendor, these new sessions offer a wealth of knowledge. If you ever wondered what the difference between a Transformer and a Transformer XL is, check out their channel! /25 youtube.com/playlist?list=…
Most people know @robertskmiles from his appearances on @computer_phile, but he also has his own YouTube channel! Take a look, if you you want to learn more about concrete problems in AI safety /26 youtube.com/playlist?list=…
When it comes to blogs and similar mediums, @gradientpub is doing an incredible job writing "accessible and technically informed" content about a diverse set of subjects in ML, from vision to language to ethics /28 thegradient.pub
A blog that I would highly recommend is @JAlammar's. His posts are some of the clearest and nicely visualized I've ever seen. Definitely give his explanations of Word2Vec, BERT and the Transformer a read! /30
• • •
Missing some Tweet in this thread? You can try to
force a refresh
It sets the stage for how to think like an attacker, and explains various types of attacks
While it focuses on ETH, a lot of concepts carry over to Solana
Then head to @osec_io’s blog to read their delineation of Solana from an auditor’s perspective.
The way they introduce Solana's Execution and Programming Model bottoms-up is helpful to get an understanding of where vulnerabilities can originate from
@anchorlang for context, Anchor assigns each `#[account]` a type with a unique 8-byte identifier and prepends it to the account data - the so-called "discriminator"
this prevents account confusions, which can lead to vulnerabilities as detailed in Neodyme's blog
Everyone has heard about fast.ai or CS231n (for a good reason), but did you know you can access Stanford’s CS224w ML with Graphs or download the book Elements of Causal Inference for free? Thread on underappreciated ML resources 📚🎥 that deserve more love 👇 /1
With the # of (new) resources online and inspired by @SEBruder, I'm trying out a newsletter to curate these resources. Next edition will be on topics like NLP, RL and GOFAI. Feel free to check it out and lmk what you think! /2 getrevue.co/profile/openml…
Stanford’s @stats385 has a myriad of fascinating lectures on theoretical deep learning: from robustness to overparameterization of NNs to DL through random matrix theory. It's a shame most of these fantastic lectures only have a few hundred views /3 stats385.github.io/lecture_videos
If you're bored at home or just want to learn something new, check out these 15 amazing reading groups/virtual seminars covering topics such as Deep Learning, Computer Vision, Robotics, RL theory, NLP and Continual Learning 🖥️🎥👇
1/ MIT's Embodied Intelligence group has a seminar series moderated by @FerranAlet that regularly invites PIs to present their research. The last talk was given by @wellingmax who talked about his group's work on Equivariant Mesh CNNs and Factor Graph NNs
2/ Another one coming out of MIT is the vision seminar organized by @xavierpuigf. Just watched @dimadamen's talk about fine-grained object interactions. Super clever how the epic kitchen dataset was annotated via narration!