I'll be binge-watching @jeremyphoward and @GuggerSylvain's @fastdotai "Practical Deep Learning for Coders" course today and tomorrow! 8 lectures, ~2h each. It's going to be fun! 😂 Why?
Well:
1/
Well:
1/
* I want to update my blog on getting started with ML from 2019 where I only recommended @coursera (and I realized just how bad my writing was just 2.5 years ago!).
The blog is here: gordicaleksa.medium.com/get-started-wi…
2/
The blog is here: gordicaleksa.medium.com/get-started-wi…
2/
I recommend you bookmark it but don't read it just yet, it should be ready by the end of this week!
* I want to be able to give better advice to "younger folks" in general. I get a lot of questions on my Discord as well (join it if you haven't: discord.gg/peBrCpheKE).
3/
* I want to be able to give better advice to "younger folks" in general. I get a lot of questions on my Discord as well (join it if you haven't: discord.gg/peBrCpheKE).
3/
Lots of smart people there!
IMPORTANT:
This is not the way to go through the course if your goal is to learn ML. I'm just skimming the course to get a feeling for what is covered and how. I'll be taking summaries along the way to improve my retention though.
4/
IMPORTANT:
This is not the way to go through the course if your goal is to learn ML. I'm just skimming the course to get a feeling for what is covered and how. I'll be taking summaries along the way to improve my retention though.
4/
You should be going through the notebooks, asking questions on the forum, and in general spacing the lectures out (temporally) in order to learn more effectively.
Check out this awesome video BEFORE doing the course (the so-called "0th lecture"):
5/
Check out this awesome video BEFORE doing the course (the so-called "0th lecture"):
5/
I'll be sharing my summaries on this thread as I progress! And finally, I'll update my getting started blog with new info.
6/
6/
Lecture 1 summary:
Lecture video: course.fast.ai/videos/?lesson…
Jupyter: github.com/fastai/fastboo…
My lecture notes:
* There is a book that goes along with the course. There is a free version (github.com/fastai/fastbook), but if you can buy the book to support them!
L1/
Lecture video: course.fast.ai/videos/?lesson…
Jupyter: github.com/fastai/fastboo…
My lecture notes:
* There is a book that goes along with the course. There is a free version (github.com/fastai/fastbook), but if you can buy the book to support them!
L1/
* There is no need for a math PhD, just get started!
* They give some historical context.
* They teach the course using the top-down approach (shown to be a much better approach for most people).
L2/
* They give some historical context.
* They teach the course using the top-down approach (shown to be a much better approach for most people).
L2/
Check out this blog of theirs fast.ai/2016/10/08/tea… to learn more about their teaching methodology.
* The stack behind the course: fastai-PyTorch-Python + Jupyter, but as Jeremy Howard says - it doesn't matter that much. Focus on learning the concepts.
L3/
* The stack behind the course: fastai-PyTorch-Python + Jupyter, but as Jeremy Howard says - it doesn't matter that much. Focus on learning the concepts.
L3/
* A high-level overview of what ML is, what SGD is, etc.
General notes:
* The course seems to require more self-discipline compared to Coursera's course.
L4/
General notes:
* The course seems to require more self-discipline compared to Coursera's course.
L4/
After the lecture, you need to go through the Jupyter, start Colab or a GPU server, play with code a bit, search through the docs. All in all, get comfortable before taking video lecture 2. Use forums to find peers (and check whether somebody created a Discord).
L5/
L5/
* Do the questionnaire, track what you didn't know in a todo list, and later get back to it!
* The code may have a weird coding style - it's a result of Jeremy mixing his understanding of various programming languages (like APL, etc.).
L6/
* The code may have a weird coding style - it's a result of Jeremy mixing his understanding of various programming languages (like APL, etc.).
L6/
Lecture 2 summary:
Lecture video: course.fast.ai/videos/?lesson…
Jupyter: github.com/fastai/fastboo…
My lecture notes:
* Getting deeper into ML theory (why do we need validation/test sets, why does transfer learning work, diff between loss vs metric, etc.)
L21/
Lecture video: course.fast.ai/videos/?lesson…
Jupyter: github.com/fastai/fastboo…
My lecture notes:
* Getting deeper into ML theory (why do we need validation/test sets, why does transfer learning work, diff between loss vs metric, etc.)
L21/
* Confidence intervals, p-value, etc. On a concrete example of whether COVID infections correlate with humidity & temperature.
* Going to production! Bear detector example (very funny). Shows you how you can create your own dataset (e.g. scrap data via Bing API).
L22/
* Going to production! Bear detector example (very funny). Shows you how you can create your own dataset (e.g. scrap data via Bing API).
L22/
General notes:
* The part on p-value may be very very overwhelming to beginners. Be comfortable with not understanding everything (that's hard if you only ever went through the formal educational system).
L23/
* The part on p-value may be very very overwhelming to beginners. Be comfortable with not understanding everything (that's hard if you only ever went through the formal educational system).
L23/
* In general younger folks (high-school kids, students) should probably take the @coursera route, IMHO. More experienced people who want to switch careers may find @fastdotai's course a better option.
* One part of the lecture was more about COVID than about ML per se 😅
L24/
* One part of the lecture was more about COVID than about ML per se 😅
L24/
* Videos are there only to help you better understand the notebooks but he also gives nice hands-on examples - so don't skip the videos.
End of Lecture 2 notes.
L25/
End of Lecture 2 notes.
L25/
Lecture 3 - summaries for other videos will be in this LinkedIn post:
linkedin.com/feed/update/ur…
So as to avoid spamming people here. I tried this as an experiment but Twitter is definitely not the best platform for this. 😅
linkedin.com/feed/update/ur…
So as to avoid spamming people here. I tried this as an experiment but Twitter is definitely not the best platform for this. 😅
• • •
Missing some Tweet in this thread? You can try to
force a refresh
