This repositories contains the material for 4 lectures given at the Italian university of applied science in Switzerland.
A lot of material on Deep Learning and nueral network can be found at my book companion website. My book (Applied Deep Learning with TensorFlow 2 published by Springer/APRESS) can be bought here.
In this lecture we will look first at Jupyter notebooks. Those are the common tool used by data scientists and machine learning engineering during the exploratory phase of the machine learning pipeline. We will look at Google Colab, an online version of Jupyter notebooks that does not require any local installation. We will use this tool during our lectures. After that we will look at TensorFlow and Keras and how they work. We will look at very basic examples leaving advanced usage to further study. Finally we will have a hands-on session in which we will use Keras to build a very easy neural network to classify images.
- Google colab - An Introduction
- Keras and TensorFlow - An Introduction PowerPoint Download Link
- Hands-on session - JUPYTER NOTEBOOK
The hands-on session will be based mostly on one Jupyter Notebook.
-
A nice video introduction to Google Colab can be found in this video. The official website for the jupyter notebook project can be found here.
-
You can also find an introduction to Neural Networks in the slides for a lecture given by Umberto Michelucci at the ETH university in Zürich, Switzerland. They are google slides and should be accessible online easily.
In this lecture we will look at what autoencoders are and what you can do with those special neural network architecture. A complete introduction can be found in this paper. We will look at the following topics
- What are autoencoders (Introduction Paper)
- Denoising autoencoders
- Autoencoders for dimensionality reduction
- Your first autoencoder
- Anomaly Detection with autoencoders
- Denoising Autoencoders with Feed Forward Neural Networks
- Denoising Autoencoders with Convolutional Neural Networks
A lot of code is available for GANs.
WARNING: the code is slow and on a normal laptop may take several hours. So you are warned. In general you can download the jupyter notebooks and use them on google colab, but also there it will get at least 4 hours to get some decent results. The examples that we will run (and the time you will see in the notebook are coming from a professional deep learning machine with 3 RTX A6000 professional graphic cards).
The presentation can be found HERE. No code examples are available for this lecture.