Introduction
Readers, the original intention of this article is to record some gadgets and blog posts that can greatly improve productivity encountered in scientific research and life. At least for now, they are all gadgets. It will be added iteratively, and it will probably be a more useful thing when it has to be sorted out to a certain extent.
text
- https://d2l.ai/index.html "Dive into deep learning" is not very clear in some places, but fortunately there are mathematical formulas.
- zhuanlan.zhihu.com/p/410776234 This is an article explaining self-attention, with clear descriptions and rich pictures, and it belongs to the level of potato chips in terms of eating difficulty
- blog.floydhub.com/attention-mechanism/ explains the application of attention in the seq2seq task, and explains two classic methods: Bahdanau and Luong. I have to say that this kind of detailed blog in foreign countries is still well written, and there are no difficult words in the whole article, which is completely different from those economics papers, which are painful to read
- https://www.joshbelanich.com/multi-headed-attention Multi-headed self-attention
- https://zhuanlan.zhihu.com/p/102697821 nn.DataParallel
- https://zhuanlan.zhihu.com/p/30844905 Introduction to RNN
- Use the plugin TamperMonkey to quickly download arxiv papers
- Plug-in download site crx4chrome.com/
- Script source zhuanlan.zhihu.com/p/36940772
- Website d.serctl.com/ for quick download of github resources
- Python re tutorial https://realpython.com/regex-python/
10. Central limit theorem https://www.scribbr.com/knowledge-base/
Make complaints
Those who put forward all kinds of seemingly advanced academic terms in the papers that people don't understand should be hanged on a tree and executed first.
MLP (Multilayer Perceptron) is a fully connected layer. Fully connected is more intuitive and beautiful. It is really...