This is really incorrect advice in the world of software engineering. At least in FAANG companies and big name startups, no one cares if you read How to Win Friends and Influence people and have "social skills". If you can't pass the engineering interviews, you're fucked.
I really hate this article. Pretty much everything I've ever accomplished in my life is because I speak it into existence and hold myself accountable. But then again, if you're the type of person who makes things happen, you don't rely on crappy psychology studies to tell you how to live your life.
I wonder why internet commentators who haven't made billions always bring up this index fund stuff, when no one who actually has made billions did it by piling their money into the S&P 500.
I took a linear algebra course. Maybe I need to restudy but I would rather have links to exact chapters than reading the whole thing, as I don't see a direct path from the linear algebra I studied to the topology concepts and geometry here. Specifically there weren't any curves in linear algebra. Ex 4.10 and 4.11 here: http://planning.cs.uiuc.edu/node143.html
Are examples of things that are hard for me to understand. Especially that 2d picture in 4.11
The topology part of these textbooks is the hardest for me. Understanding the 2d representations of the shapes. Where can one get a more bottom up explanation of the crucial topology concepts? Also bad at geometry in general
In order to get a variable length context, you need to add some machinery to some forms of attention. For example, in jointly learning to align and translate, the attention is certainly not invariant to number of context vectors. You train the attention to take in a fixed number of context vectors and produce a distribution over the fixed number of context vectors. You cannot train on images with 5 annotations/context vectors and expect anything to transfer to a setting with 10 annotations. That's why I would be interested in a specific paper to solidify what you're saying.
Samcodes said it above. How do transformers build a shared representation of two input sentences with different lengths? If you convolve them with the same filter, you get two different sized convolution outputs - the embedding dimensions don't align.
How do you handle variable length input without something like an RNN? Even transformers use RNN structures right.
I suppose convolutions could technically handle variable length inputs (just slide the window of weights over different length inputs) but I don't think tensorflow or pytorch supports this