Skip to content

Advanced Machine Learning Workshop: From Theory to Practice with Neural Networks

License

Notifications You must be signed in to change notification settings

RISCSoftware/dl-part2-workshop

Repository files navigation

Abstract

Advanced Machine Learning Workshop: From Theory to Practice with Neural Networks

We cordially invite you to participate in a comprehensive workshop designed to enrich your understanding of machine learning. We will explore the theoretical foundations and practical applications of neural networks, emphasizing the transformative technologies of transformers and attention mechanisms.

Throughout this three-hour workshop, we will explore both theory and hands-on coding, using Python and Jupyter notebooks to apply learning in real-time. We will cover essential aspects such as neural network architecture, emphasizing initialization, regularization, optimization, backpropagation, and first-order automatic gradient differentiation. These concepts form the backbone of advanced machine learning practices. In our session on natural language processing (NLP), we will discuss and code the attention mechanism and transformer architectures, combining theoretical insights with practical coding exercises. The workshop will also include discussions and examples on select papers in the field of neurosymbolic AI, exploring current trends and innovative methodologies.

Please bring your laptop to fully participate, as each session integrates coding activities where we implement and train key components of neural networks using PyTorch. Detailed instructions for preparing your laptop, including necessary installations and configurations, will be provided soon. We give our best to offer a workshop that will enhance your understanding of neural networks and provide practical experience with advanced machine learning techniques for tackling complex challenges.

How to run the workshop code?

Option 1: Run a docker container

  • Install docker
  • Start our docker container
    docker run -p 8888:8888 risclidse/dl-workshop bash -c "jupyter notebook --NotebookApp.ip=0.0.0.0 --NotebookApp.port=8888 --NotebookApp.allow_root=True --NotebookApp.notebook_dir=/repo --NotebookApp.token=''"
    
  • Enter the provided URL in the browser, e.g. http://127.0.0.1:8888/tree

Option 2: Run the notebooks on Google Colab

  • Download the .ipynb files of our workshop
  • Go to https://colab.research.google.com/
  • Select File → Open notebook → Upload
  • Note:
    • You might be asked to restart the session once after the pip3 install completed
    • Most images will not be shown on Colab

Participate in Live-Coding

Welcome to the workshop! We're excited to have you join us for this interactive learning experience. To help you get started smoothly, here's everything you need to know about setting up your environment.

At the start of the workshop, you will receive a unique ID between 1 and 30. In the following instructions, <id> will represent your specific ID. You can already test your access by using any ID. Feel free to run code or modify files—everything will be reset before the workshop begins.

Please be aware that we are currently experiencing issues with the VSCode option. If you encounter an error message stating that VSCode cannot execute Docker commands, please try again later.

If you see only a blank page in the Browser option, try Ctrl+Shift+R.

Option 1: Browser

Option 2: VSCode

  • Install or update VSCode on your laptop
    • Via an executable or using a command, e.g. in Windows:
      winget install -e --id Microsoft.VisualStudioCode
      winget update -e --id Microsoft.VisualStudioCode
      
      In Debian Linux:
      sudo apt update
      sudo apt install code
      sudo apt upgrade code
      
  • Open VSCode and install extensions
    • Sidebar → Extensions → <enter name> → Install
      • Remote Development
        ms-vscode-remote.vscode-remote-extensionpack
      • Python
        ms-python.python
      • Jupyter
        ms-toolsai.jupyter
  • Connect via SSH
    • Ctrl+Shift+P → Remote-SSH: Connect to Host...
      • Enter <user>@qftquad2.risc.jku.at
        Note: Make sure to prefix your user on Windows
      • Enter your password
    • Ctrl+Shift+P → Dev Containers: Attach to Running Container...
      • Select dl-workshop-runtime-<id>
      • Enter your password again
  • Sidebar → Explorer
  • Open Filder → /repo/
  • Navigate to workshop/1a_torch_tensors.ipynb

File structure

Attribution and Licence

This project uses content from the following sources:

Please refer to the links for specific license details.

About

Advanced Machine Learning Workshop: From Theory to Practice with Neural Networks

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published