Beginner Intro to Neural Networks 12: Neural Network in Python from Scratch




[ad_1]

Handwriting generation with recurrent neural networks: https://www.cs.toronto.edu/~graves/handwriting.html

Notebook with neural net:
https://github.com/JonComo/flowers/blob/master/flowers.ipynb

Music:
Pookatori and Friends Kevin MacLeod (incompetech.com)
Licensed under Creative Commons: By Attribution 3.0 License
http://creativecommons.org/licenses/by/3.0/

Hey everyone!

In this video we solve the flower problem “by hand” using python, and do it interactively in jupyter notebook/IPython.

You’ll see everything from weight initialization, to the feed forward and backward passes used to train the net, to inference on some crazy types of flowers.

Also as a bonus I left in my hyperparameter (learning rate, training iterations, weight variance etc.) “dancing” for fun!

Thanks for watching. Appreciate seeing you all here still. I’m starting to work on these videos full time. Let me know what you’d like to see next!

See you in the next video,
– gnn

Source


[ad_2]

Comment List

  • giant_neural_network
    November 16, 2020

    is this channel DeD?

  • giant_neural_network
    November 16, 2020

    Can I add more features and more types (outputs) of the flowers to do the excect same thing? Without sigmoid function of cause

  • giant_neural_network
    November 16, 2020

    Hello from the other side! Really cool visualizations, and I may or may not have taken notes to include you in a machine learning report/project I'm doing that tries to summarize concepts that are lacking from traditional ml courses

  • giant_neural_network
    November 16, 2020

    hey great video, just wanna ask about the TensorFlow video that you talked about at the end?

  • giant_neural_network
    November 16, 2020

    Super easy to follow and brilliant testing at the end was super cool to test all the weird combinations. Did you ever manage to do this in tensor flow?

  • giant_neural_network
    November 16, 2020

    missing u this series bro

  • giant_neural_network
    November 16, 2020

    When ever I go to print anything in my for loop it gives me error( numpy.float64 object is not callable)….can sombody please help

  • giant_neural_network
    November 16, 2020

    this entire series was a giant waste of time, i made sure to dislike the video.

  • giant_neural_network
    November 16, 2020

    Do you cover backpropagation in this series?

  • giant_neural_network
    November 16, 2020

    Hi, I followed you correctly, but my cost_sum is going in the opposite direction of minimisation. Now its graph is forming a Nike symbol, i.e., maximising

  • giant_neural_network
    November 16, 2020

    Hi, i am trying to develop a python neural network module for learning purpose thanks to your really amazing videos, and after trying to figure out everything, i came up to a point where everything seemed correct but… something strange is happening in my train method. My cost value over every epoch does not go lower than 1. I would really like to correct any mistake i did but i do not know what it is :/ Would you help me to figure it out? I would really appreciate it!

  • giant_neural_network
    November 16, 2020

    Shouldn't you be re-inistializing the target variable in your little costs_sum loop?

  • giant_neural_network
    November 16, 2020

    Hey… Thx for the tutorial man. It was very easy and helpful. But here's the thing while iterating for 10000 times it works fine however when say 50000 its accuracy suffers significantly. Is this a common thing for NN or is this only for this dataset. I see in the video you had a similar problem. Please reply man.

  • giant_neural_network
    November 16, 2020

    Hands down the best possible series. I was jumping between tutorials as I couldn't understand the other ones, this is the best and precise yet, amazingly simple series ever. This just gave me a push in the right direction that I needed.

    Thanks a lot for the efforts you put in into the videos and all the explanations. You got me even more interested in neural networks and even maths that I absolutely hate. Man, I love you and I owe you a big one. Once again thanks a lot.

    Hope you keep making such series and gain many more subscribers so that people know how great educator you are.

  • giant_neural_network
    November 16, 2020

    Are you going to proceed your multi-neural network series? Please doo!!!

  • giant_neural_network
    November 16, 2020

    Setup some Patreon and we would gladly join!

  • giant_neural_network
    November 16, 2020

    This is BY FAR the best intro to NN! Thank you very much! Please, keep going!

  • giant_neural_network
    November 16, 2020

    I am always picky about liking videos and even commenting on them. I loved this series from the beginning but was still in doubt until I came to the tenth video. You just made me login to my account and like each one of your video in this series and comment on them. That’s how good this series was. Thank you 😀 (Warning : Since I wrote the comments sometime after watching the video, some of them might be cringey)

  • giant_neural_network
    November 16, 2020

    Thank you sm!

  • giant_neural_network
    November 16, 2020

    I dont normally comment but man, forreal? I agree completely with Tejas Arlimatti this is how I learned the breakdown of a neural network. Now it's been 2 yrs since this video I would really like to see the break down of more. How to add a few hidden layers to this? How to take it to the next step build a CNN, RNN, use ARS, Activation function ReLU?

Write a comment