Hand-in assignment 1

Implementation of a forward neural network and training with mini-batch gradient descent.

Important: The implementation should be done by implementing the relevant equations in numpy/Python or similar, not using high level languages like Tensorflow/PyPorch.

• Instructions: hand-in-assignment-1.pdf Download hand-in-assignment-1.pdf
  (only updated with new deadline since previous version)
• Data: MNIST.zip
• Code for loading data: load_mnist.py Download load_mnist.py Links to an external site. (python)
• Info about data
• Due date: April 19 (new deadline)
  
• Due date for peer-review: April 22 (new deadline)
  

If you have questions you can ask them here.

Comment on submission format: You can choose if you want to add the code as an appendix in the pdf or in a separate zip-file. If you choose to submit the code in a zip-file, the pdf must be outside the zip-file. Hence, if you go for the zip-alternative, you should submit two files, one pdf and one zip-file with just the code. If not, neither the peer-reviewers nor the teachers will be able to see and comment in the pdf via the speedgrader tool. If you go with the zip-alternative, it is also important that you submit pdf and zip in the same submission (not in two separate sequential submissions), since the peer-reviewer will only see your most recent submission.