llsourcell/how_to_do_math_for_deep_learning

This is the code for "How to Do Math Easily - Intro to Deep Learning #4' by Siraj Raval on YouTube

how_to_do_math_for_deep_learning

This is the code for "How to Do Math Easily - Intro to Deep Learning #4' by Siraj Raval on YouTube

Overview

This is the code for this video on Youtube by Siraj Raval apart of the 'Intro to Deep Learning' Udacity nanodegree course. We build a 3 layer feedforward neural network trains on a set of binary number input data and predict the binary number output.

Dependencies

None!

Install Jupyter notebook from here

Usage

You can either run the notebook by typing jupyter notebook into terminal when in the directory or run the demo.py script by running python demo.py in terminal.

Weekly Challenge

The challenge for this video is to build a neural network to predict the magnitude of an Earthquake given the date, time, Latitude, and Longitude as features. This is the dataset. Optimize at least 1 hyperparameter using Random Search. See this example for more information.

You can use any library you like, bonus points are given if you do this using only numpy.

Due Date: Thursday, February 9th at 12 PM PST

Credits

Credits for the original code go to Andrew Trask. I've merely created a wrapper to get people started.

how_to_do_math_for_deep_learning's Issues

Syntax error in python3.

anyone knows what's the exact problem here?
it's giving me a syntax error in python3

-----k2_delta = k2_error * nonlin(k2,deriv=True)
------------^
SyntaxError: invalid syntax

Error in forward prop

In line 28, what is written is:
k0 = X

It should be:
k0 = nonlin(X)

this is so that the weights are effected by the output of the node and not the input, and would also be consistent with the k1 and k2 values used.

Can you explain the last two lines where the gradients are being updated?

syn1 += k1.T.dot(k2_delta) syn0 += k0.T.dot(k1_delta)
Could you explain a little more in depth the thought process behind these lines? Came here from the video and still confused. How does k1.dot k2_delta create the amount to update the weight? I understand the gradient is the direction we need to travel but wouldn't we just subtract that distance from the weight? Why do we need to dot product?

Thanks!

Syntax Error in demo.ipynb

When I run the Jupiter notebook, line 14 will not display unless the variable is in parentheses. Anyone else encountering this?