Neural Networks from Scratch - P.9 Introducing Optimization and derivatives

Introducing the challenge of optimization and the concepts of derivatives

#deeplearning #machinelearning #nnfs

[adityataggar]

At 8:44 shouldn't it be y2-y1 instead of y1-y1?

[ProGamer]

Top three unfinished works of all time:

3. The Silmarillion by J. R. R. Tolkien
2. Requiem by Mozart
1. "Neural Networks from Scratch" YouTube series by sentdex

[redthunder]

please do a part 10, this is the only series that I found on youtube that actually explains the concepts. and its done so well

[BollForte]

"and I will see you in the next video ...." never was so disappointed :(

[Aryan-fdyv]

please complete this series it is literally the best one i've ever seen on the internet

[Skybot]

Is part 10 happening? This has been the best series I've been able to find for this sort of thing.

[mechyrexic]

Oh my god I just discovered this series today and I was bummed the optimization videos weren’t out yet, and of course, what perfect timing for this video to come out an hour after I finished episode 8, thanks so much!

[lathryx]

I've just finished watching all 9 of the videos currently on this playlist. I absolutely cannot wait for part 10. PLEASE CONTINUE.

[MoonberryJam]

This series is absolutely amazing. Your explanations are extremely thorough while still being engaging - something that even my favorite college professors have trouble with. I feel like I'm actually learning ML instead of just copying and pasting code. Can't wait for the next video!

[marcoflo]

Start an amazing series based on your book. Leave out the conclusion of the problem. Well played..

[judedavis]

Thanks for making these. It must be a lot of hard work, we all appreciate it so much. People say that ‘reinventing the wheel is a waste of time’, but learning how the wheel works, what makes it tick, helps us to apply more advanced techniques. Thanks once again 🙏🏼

Man you cannot cut such an awesome content in the middle like that, pleeeease finish. I never asked you anything, please xP

[DRealHammer]

edit: just noticed, someone already mentioned this, but maybe the explanation is appreciated by someone

Cool video, just a little inaccuracy:
You mentioned at 8:16: "this can only be done on a continuous curve, so no sharp corners"

I know this is an introduction but to clarify:
a continuous function can have sharp corners. A better description for a continuous function would be that it should be " connected", so no parts, where it suddenly jumps from 1 to 0 without a connection. Someone gave an intuition like "if you can draw it without lifting your pencil it's continuous. But there's a bit more to it.

what we want to look at are "differentiable function", so functions you can differentiate (or find a derivative).
Does not help with understanding, but these are the functions with no sharp corners.
All differentiable functions are continuous. So it is a stronger statement.

Examples:
f(x) = x^2 is differentiable and hence also continuous

f(x) = abs(x) is continuous, but no differentiable (sharp corner at 0, as the function describes a V shape)

f(x) = 1, if 0 < x < 1; else 0 is not continuous, hence also not differentiable (we have value jumps at x = 0 and x = 1)

[jameshughes]

I'm so glad to see this series continued. This is a gold mine of a lesson.

[ParametricCPA]

This series has blown my mind so far! I went from 0-60 real quick, but I feel like I understand it! Can't wait for part 10!

[raccoon_]

This is the best tutorial series I've found on this subject anywhere. Followed upto this point with no issues. Hopefully you'll continue this someday :(
Thanks so much for the vids so far 👍

[harshalchalke]

A very gifted teacher and great content creator. Finally found a teacher who explained the concept so well, but it's a shame that he settled for being a salesman. Was about to recommend this series to my class, but how can I? Left at the most important part. Thanks though, you explained it very well. I hope you decide to complete the series for future learners.

[splch]

this series is seriously a masterpiece. the visuals, information, and presentation are just incredible :) i cant wait to see how this series is used in the future

[polymer]

hope to see this series brought back to life, I bought the book first and really like having the videos to confirm what I read.

[muhammadfarhanbinphakhrudd]

Absolutely learned alot from this series, not just for my technical knowledge in neural networks but also my calculus subjects seems to be getting better. Knowing that the knowledge can be applied.