Here are my two cents.
The list of math topics suggested by Fawaz abdul rahman looks good and
pretty comprehensive. Since no one else mentioned accessible courses, I
typically go to wikipedia - and I found it the only resource that is
always accessible. You can either browse source code of the pages - but
you would need to understand Latex notation for that; or figure out how
to read formulas right there on the page - if you use NVDA you can use
excellent Access8Math add-on which works great for wiki formulas. The
problem with wikipedia however is that it is not structured as a course,
so you would need to figure out the order in which to read those
articles and also you would need to figure out which parts of the
articles to skip, since they often bundle together simple concepts with
really advanced stuff, that most people wouldn't need.
As for ML topics, again, since no one has mentioned, let me try to come
up with a list. I don't know any accessible course, but again what you
need can probably be all learnt from wikipedia. Here is the list:
1. Linear and logistic regression.
2. Basic neural networks
3. Backpropagation
4. Metrics (L1, L2 distance, Cross Entropy, Precision/Recall)
5. Overfitting/underfitting
6. Regularization/normalization
7. Cross-validation
8. Gradient-boosted decision trees.
These are the basics.But ML has evolved a lot during the last 10 years,
so here are some key inventions that you hsould get familiar with, at
least it would be beneficial if you roughly know what is it all about:
1. Deep learning
2. Deep reinforcement learning
3. Optimization algorithms - adam, adagrad
4. RNN and LSTM
5. Attention and seq2seq modeling
6. Transformer architecture and language models
I must have missed something but this list should be good enough to get
you started.
Now as for working in industry. In my experience there are many kinds of
teams that work on ML. You probably don't want to work in cutting edge
ML, since this requires reading papers and PDF format just sucks. But
the good news is that there is a lot of demand for simple blackbox ML -
that is where you use blackbox frameworks to train models without
digging much deeper into internals. And in this type of work I didn't
have much issues with accessibility. Jupyter and similar tools tend to
be pretty accessible. PyTorch and TensorFlow are just frameworks, no
accessibility problems here either. Obviously you shouldn't work in a
team that does image processing, but there are many other ML problems,
such as ranking, text processing, various kinds of classifiction that
doesn't require looking at images. And in my experience I didn't find
much need to visualize any graphic information. If you have to visualize
something, you can sonify either using hassaku's library already
mentioned above, or using my AudioChart add-on for NVDA.
HTH
--Tony
On 1/30/2022 9:28 AM, Fawaz abdul rahman wrote:
Hello,
not sure if the resources mentioned in the following article are accessible, but the list looks quite interesting.
All the Math You Need to Know in Artificial Intelligence <https://www.freecodecamp.org/news/all-the-math-you-need-in-artificial-intelligence/>
On Sun, Jan 30, 2022 at 7:23 PM hamidreza abroshan <hamidreza.abroshan@xxxxxxxxx> wrote:
Thank you Mohammad for your great notes and Hassaku for sharing
your repo and also, everyone who participated.
I'll follow your comments.
If you introduce some basic books, courses and a beginner road map
for this subject, I'll appreciate it.
And the course that you introduced for math, if it needs some
basic background, can you recommend a more basic math tutorial also?
Because I studied in an school for blind, and my field was human
science, they did not take math serious and now I have to study more.
Thanks.
On Sat, Jan 29, 2022 at 9:29 PM Muhammad Fayed
<m10fayed@xxxxxxxxx> wrote:
Great job Hassaku.
I like your work so much and starred your repositories. 😊👍
Thank you so much,
Regards,
Mohamed E. Fayed
> On 29 Jan 2022, at 5:09 PM, hassaku <hassaku.apps@xxxxxxxxx>
wrote:
>
> Hi hamidreza,
>
> Some people have already mentioned it, but if you want to
start practicing machine learning in a casual way, you can use
Google Colaboratory.
> You don't have to bother with building a troublesome
environment, and although it's not perfect, you can use it
with a screen reader.
>
> If you need graphs, the graph sonification library I'm
developing might be useful.
> The following site contains the libraries I am developing,
and links to an introduction to machine learning.
> If you are interested, try to access them.
>
> https://hassaku.github.io/DS-and-ML-with-screen-reader/
>
> It's still a work in progress. So if you have any feedback,
please feel free to let me know.
> I hope this helps you.
>
> Best regards,
>
> hassaku
> ** To leave the list, click on the immediately-following link:-
> ** [mailto:program-l-request@xxxxxxxxxxxxx?subject=unsubscribe]
> ** If this link doesn't work then send a message to:
> ** program-l-request@xxxxxxxxxxxxx
> ** and in the Subject line type
> ** unsubscribe
> ** For other list commands such as vacation mode, click on the
> ** immediately-following link:-
> ** [mailto:program-l-request@xxxxxxxxxxxxx?subject=faq]
> ** or send a message, to
> ** program-l-request@xxxxxxxxxxxxx with the Subject:- faq
** To leave the list, click on the immediately-following link:-
** [mailto:program-l-request@xxxxxxxxxxxxx?subject=unsubscribe]
** If this link doesn't work then send a message to:
** program-l-request@xxxxxxxxxxxxx
** and in the Subject line type
** unsubscribe
** For other list commands such as vacation mode, click on the
** immediately-following link:-
** [mailto:program-l-request@xxxxxxxxxxxxx?subject=faq]
** or send a message, to
** program-l-request@xxxxxxxxxxxxx with the Subject:- faq
-- hamidreza