ssshaker / nndl-solutions

Solutions of the exercises and problems from Michael Nielsen's book Neural Networks and Deep Learning: http://neuralnetworksanddeeplearning.com/

NNDL-solutions

Solutions (math and code) of the exercises and problems from Michael Nielsen's book Neural Networks And Deep Learning (and adaptations to the code for Python 3 and Theano 1.0.3).

Here's where to find the solutions to exercises and problems:

• involving math: notebooks
• involving code: implemented in code, discussed in notebooks

With links to nbviewer:

• Chapter 0: Update the code for Python 3
• Chapter 1: Using neural nets to recognize handwritten digits
• Chapter 2: How the backpropagation algorithm works
• Chapter 3: Improving the way neural networks learn
• Chapter 4: A visual proof that neural nets can compute any function
• Chapter 5: Why are deep neural networks hard to train?
• Chapter 6: Deep learning

TODO

So far, I've provided solutions to all exercises and problems, except:

• chap3 p8, p9
• chap4 p1 (c)
• chap6 p5, p7 (theano part)

Contributing

I may have made mistakes, or provided incomplete or suboptimal solutions. And there are still some problems that I haven't solved. So direct improvements, or any suggestions, are much welcome!