Part Four: Improving Deep Neural Networks

This section contains the bulk of the course. We really want you to have a broad knowledge of everything you can do to get your model performing extremely well. This includes:

  • Hyperparameters

  • Hidden layers

  • Number of Neurons

  • Initialisation Logic

  • Activation Functions and Non-Saturating Activation Functions

  • Normalisation methods

  • Momentum, Adaptive and Hybrid Optimisation methods

  • Regularisation and Monte Carlo Dropout

The course is then concluded with a final project. This workbook will get you building a neural network using all the lessons learned in the course, as well as getting you to think. We will use data from 30,000 customers that use credit cards to predict if they will default on their payments next month. We use Monte Carlo dropout to make risk adjusted predictions to give us a better risk and cost profile.​

12 - Hyperparameters, search, and architecture
13 - Vanishing Gradients
14 - Glorot and He initialisation
15 - Non-saturating activation functions
16 - Batch Normalisation and Gradient Clipping
17 - Momentum based optimisers
18 - Adaptive Learning Rate based Optimisers
19 - Regularisation and Dropout
20 - Monte Carlo Dropout
21 - Summary and Overview of Final Project
22 - Final Course Project
Programming Exercises
Final Course Project: Predicting Credit Card Defaults

Download the files below and open up the WORKBOOK to begin.

You also need to download the data seperately, which can be found here:

https://www.kaggle.com/uciml/default-of-credit-card-clients-dataset

Please place all the files in the same folder before beginning.

End of Course!

Congratulations! You have completed the course "A Practical Introduction to Artificial Neural Networks for Finance". I hope that you have learned a lot of practical knowledge that can be taken away and applied to new problems. This is by no means the end of your journey into studying deep learning. You now have the knowledge to build models and now go back into the mathematial details of how they really work. You should have a good intuitive background with which to look at such mathematics.

If you have any feedback for the course, please contact us here:

Contact Us

FinanceAI.co.uk

  • Black YouTube Icon
  • Black LinkedIn Icon