Coding a Neural Network: A Beginner's Guide (part 5)
Вставка
- Опубліковано 4 жов 2024
- Neural networks simplified and made easy, for the layperson (or medical practitioner). I've tried to keep things simple, and provide a beginner's introduction to machine learning and neural networks. By the end of this series, you'll have created your first complete and functioning artificial neural network, written in Google Colab. I recommend listening on 1.5 or 2x speed.
In part 5, we finally get our network learning. You'll see your super simple network learn about the dataset its being fed, and improve the accuracy of its predictions based on that dataset.
Part 1: • Coding a Neural Networ...
Part 2: • Coding a Neural Networ...
Part 3: • Coding a Neural Networ...
Part 4: • Coding a Neural Networ...
This tutorial is greatly indebted to the work of Justin Johnson: github.com/jcj...
Let me know what you think in the comments below👇
Loved the whole 5 parts!
Can you please apply these neural networks on real life medical problems solving? Thank you
Thank you for sharing you knowledge!
Thanks for the great tutorial.
It would be amazing if there were more videos like this for healthcare workers.
Greatly explained
Thanks so much, for this simplified basics. However, I was wondering when the biases come it?
I've know know to make a neural network framework for ages but I could never understand how to get it to "learn", but this was perfect though and I understand it now. thx.
epic tutorial
Great series. Why is bias not used in this tutorial?
outstanding
Cool series bro! 👍
Thank you!
Thank you very much for your tutorial.
as far as my understanding is the only data can be reused is o_data[]. is it right. If yet, the calculation alway do heavy job each different input data is changed.
Should we save the optimized w1[], w2[] for the next calculation instead of new random values when it run with new input
thank you
I am so confused how we actually use the trained network. Like how do we ask it to predict outputs depending on inputs we give?
it's like a function y=mx+b but a lot more fancy and complicated. Plug in the inputs (x,y,z, or whatever the inputs are) and the network will give you a predicted number (y or f(x)). basically the o_data_predictions is the numbers the network outputed.
I got an error and I am confused too much, can someone help me?
ValueError Traceback (most recent call last)
in ()
16
17 for i in range(1000):
---> 18 h_values = i_data.dot(w1)
19
20 h_relu = numpy.maximum(h_values, 0)
ValueError: shapes (8,2) and (3,2) not aligned: 2 (dim 1) != 3 (dim 0)
Edit: I checked my code and saw that at the last line, I wrote w1 = w2 - grad_w2 * 1e-4. I fixed it and error was gone.
i.data.T.dot(w1)