Do you have a good reference(s) that starts from the basics all the way to CTC loss? The distill.pub article is good, but personally doesn't provide a good enough intuition...
You can study CTC in isolation, ignoring all the HMM background. That is how CTC was also originally introduced, by mostly ignoring any of the existing HMM literature. So e.g. look at the original CTC paper. But I think the distill.pub article (https://distill.pub/2017/ctc/) is also good.
For studying HMMs, any speech recognition lecture should cover that. We teach that at RWTH Aachen University but I don't think there are public recordings. But probably you should find some other lectures online somewhere.
The relation of CTC and HMM becomes intuitively clear once you get the concept of HMMs. Often in terms of speech recognition, it is all formulated as finite state automata (FSA) (or finite state transducer (FST), or weighted FST (WFST)), and the CTC FST just looks a bit different (simpler) than the traditional HMMs, but in all cases, you can think about having states with possible transitions.
This is all mostly about the modeling. The training is more different. For CTC, you often calculate the log prob of the full sequence over all possible alignments directly, while for HMMs, people often use a fixed alignment, and calculate framewise cross entropy.
If there are only non-farmers, the farmland doesn't produce any food (discounting wild berries or whatever grows when farmland is wasted). The farm can simply produce for 1.2 people...
I agree the post makes valid points, but is there anything new in that? It had been discussed several times here and on other forums as well. "RSE" is just another made-up position with a very average pay structure -- even this is not new.
However, RSEs (or just general software training) may help research groups establish a structure on how to format code, put some standards in place, and at least have some basic tests. This way, more people can read/modify the code efficiently (more = not necessarily general public, but it at least helps incoming grad students/postdocs to pick up the project easily).
Also, for all the "projection" they talk of, they are projecting US-centric views on to the wider world innit? Not many countries have cops carrying military grade guns/tanks and randomly shooting...
>they are projecting US-centric views on to the wider world innit?
Projecting, maybe, but at the same time the last few weeks have shown there's a not insignificant amount of people in other Western countries who are addicted to American culture, latched closely to the teat, who are all too eager to have American issues imported into their own societies.
I think that's because it is taught by people who are not well trained in CS themselves for the most part, and a not so well defined curriculum. "You can build a cool website/mobile app" can help only so much in inculcating an interest in computing.
I wish there's something more like this book (https://en.wikipedia.org/wiki/Structure_and_Interpretation_o...) for schools which show that computation is a good extension of mathematical thinking (when used in the sciences), and beyond that, it can used to build "cool stuff" like mobile apps or software to control robotic arms.
In case my last sentence was not clear, I'm trying to allude to different roles of software. One can use it as a useful, very fast but dumb calculator doing computations to predict a molecular structure or a spreadsheet doing a bunch of stuff for an accountant/finance guy. Many times, we tend to think of it as useful to build cool tech apps/websites and force feed them Java/C++ or some such things which they're not ready for yet[1]. We should inculcate "computational thinking" before we teach them Javascript websites/java apps.
[1]As per my experience with some fancy schools teaching 5th/6th grade kids how to build their first apps
I'm currently trying to teach a friend structured software design via the https://deinprogramm.de [0] approach based on Racket/Scheme, resisting a friends opinion on it being not useful enough in industry and all that.
The thing is, that while I'll have an easy time finding her a job with no real experience if she can demonstrate a structured approach I can vouch for, the opposite is far harder. If she can write python, fine. But no one will pay her/teach her without a long trade school contract if she can't demonstrate that she can be left alone with a task that takes 50~150 hours. Or at least that she only lacks a bit experience and needs some guidance to end up with a concise overall design.
Considering that the most senior engineers spend far more time on high-level design than on coding, it seems the better carreer path.
The book you search might be the one linked below, btw.
ok, extremely basic is a bit oversimplifying it. When you start reading quantum algorithms, you will inevitably come across Shor's factorization algorithm, which requires (quantum) phase estimation: https://en.wikipedia.org/wiki/Quantum_phase_estimation_algor... which requires quantum Fourier transform and some good deal of math. This is when you don't go into the physical implementations. If you want to look at that aspect, things may become a bit more complex.
This is not to discourage anyone, but underselling it as requiring elementary linear algebra is not very helpful (the pop-sci articles have already been overselling it as "magical"/"mind-blowing" etc.).
There are algorithms which involve advanced mathematics on classical computers, too. You don't have to understand them to understand how classical computers work. I've never bothered to learn the general number field sieve, and similarly I've never bothered to learn Shor's.
I say if you understand gates as unitary matrix multiplication, representing multiple qbits with the tensor product, entanglement, and projective measurement, you basically understand quantum computing. Throw in an algorithm or two to convince yourself of the benefits.
True, but:
1. Any serious quantum computing course/book will have Shor's algorithm in the first few chapters (in fact there are not a lot of quantum algorithms which have clear advantage over classical ones). One can teach quite a bit of useful classical algos (sort, binary search, tree/graph-based) without going into mathematics like FFT or jpeg coding.
2. Again valid, but IMHO measurements (and PoVMs) can lead to deep rabbit holes, and I found myself digging in much deeper.
Probably I should read easier expositions to see how effectively they teach. (I come from a EE+physics background, so I do gravitate to math-heavy rigorous explanations)
actually I would like that for my running shoes, I live in fear that changing shoe model would re-trigger knee issues, and I almost would like to always buy the same shoes.
