Programming Machine Learning

Paolo Perrotta Programming Machine Learning

Hands on introduction to machine learning with focus on supervised learning and the deep learning “explosion” after 2012

• CNN - Convolutional Neural Network (2012) - perfect for image recognition
• GAN - Generative Adversarial Network (2014)- for image creation
• RNN - Recurrent Neural Network - perfect for processing of language and sentences where “context” / “memory” is required

OpenAI / ChatGPT / Transformers

• ChatGPT is everywhere. Here’s where it came from
  • traditionally language models have been using RNN (recurrent neural networks)
  • in 2017 a group of Google researchers invented transformers - described in Attention Is All You Need
• How ChatGPT actually works
• How does ChatGPT work? Tracing the evolution of AIGC

Software

• Jupyter - machine learning notebooks with a mix of text, live code, visualizations, etc.
• Keras simple API on top of Tensorflow
• Tensorflow - machine learning library developed by Google
• PyTorch - machine learning library developed by Facebook
• Teachable Machine - can be used for quick’n’dirty online training of a model that can be exported to TensorFlow
• scikit-learn - an alternative open source ML library. The documentation about Choosing the right estimator contains a very informative interactive “cheat-sheet”

Hardware

• Coral - Google’s offering of small IoT components with AI support out-of-the box.

Setup using Miniconda

• install Miniconda, a minimal Conda distribution
• create an environment conda create --name=machinelearning python=3
• install packages

  conda install numpy=1.15.2
  conda install matplotlib=3.1.2
  conda install seaborn=0.9.0
  conda install scikit-learn=0.22.1
  conda install keras=2.2.4
  conda install jupyter==1.0.0
  

Activate Environment

conda update -n base -c defaults conda
conda activate machinelearning

Start jupyter Notebook

jupyter notebook