Making the first Layer in Go | Let's learn - Neural networks from scratch in Go - 4

Подписаться 847

50% 1

Today I refactored the code from previous video and implemented batch processing of inputs and neurons. Now it's almost ready to make real neural network with several layers
We study the book nnfs.io and implementing all the code with Golang instead of Python. The aim is to learn how neural networks are made inside and how it looks like to build is fully from the scratch without using frameworks like pytorch or tensorflow. We’ll have to find alternative in Go or reimplement some operations fully from the ground.
🖥️ You can support my channel, subscribe for more content or write directly to me: ko-fi.com/roman_v
📹 Main channel: bit.ly/RomanV
✉️ Contact: roman.volkov.code@gmail.com
👀 Disclaimer:
Views and opinions are my own and do not represent or reflect the opinions of my current or past employer(s).
`#learncoding #softwareengineering #neuralnetwork #machinelearning #ai`

Опубликовано:

20 сен 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 12

@_Roman_V_Code 4 месяца назад

🖥 You can support my channel, subscribe for more content or write directly to me: ko-fi.com/roman_v 📹 Main channel: bit.ly/RomanV

@Unaimend 2 месяца назад

That chapter was a little tricky :) Got a little confused if I have to transpose the output of my first layer or not (at least I had to do that in my implementation because I don't transpose my input weights). Not sure if that matters much down the line

@_Roman_V_Code 2 месяца назад

Yeah, I know what you mean. Later you will see that I had issues because of this too as I got confused where I used Transposition and where I did not. Then I decided to go every where in one style and it become easier

@_Roman_V_Code 2 месяца назад

I think it will be in 19 and 20 episode

@Unaimend 2 месяца назад

Now following along in rust, let's see whose code will be faster 🤣 But this will be my first project in rust so probably yours 😅

@_Roman_V_Code 2 месяца назад

well, that's an interesting question. Not sure that much. Gonum uses BLAS for matrix multiplications and it very optimized pkg.go.dev/gonum.org/v1/gonum/blas/blas64

@Unaimend 2 месяца назад

@@_Roman_V_Code Ohh well, then definitely yours. I am still working with whatever a `vec` is in Rust... so no BLAS for me.

@_Roman_V_Code 2 месяца назад

But my implementation is far from ideal. But let’s see how it goes))

@zacharyendrulat 3 месяца назад

sorry to ask this but are you able to link the code?

@_Roman_V_Code 3 месяца назад

Hey! I did not publish the code, but as you asked: github.com/RomanVolkov/nnfs_go It contains current state of the code from episode 27 and I will continue to push the code after each video

@zacharyendrulat 3 месяца назад

@@_Roman_V_Code thank you thank you! your videos are really good and I hope you keep going with them. I am learning a lot from them.

@_Roman_V_Code 3 месяца назад

@@zacharyendrulat Music to my ears! thanks a lot, I will keep going for sure :) Btw, you could check my main channel: www.youtube.com/@_RomanV_ Maybe you will find something useful there too