AlphaZero from Scratch
Machine learning course published on freeCodeCamp.org. Here, I rebuild the reinforcement learning algorithm AlphaZero developed by DeepMind. It can learn how to play games like Chess, Shogi, and Go at a superhuman level just by playing against itself.
Use the movements of your thumb to play Pong against a pre-trained Double-DQN-Agent. I used Google-Colab for training the pyTorch model and created the Pong environment with Pygame.
Streamlit web application that uses trained models (e.g. CNNs) to classify digits drawn by users or generated by a Conditional-WGAN-GP. One can choose between models from Pytorch, Keras, and Scikit-learn.
Stock forum with live data, created with Python (Django) and JavaScript. The data comes from the Yahoo-Finance-Api, and users have the ability to create accounts and exchange information with others about specific stocks. Talkstocks
I trained a Generative Adversarial Network (GAN) using 10.000 images of CryptoPunks. In order to prevent mode collapse, I implemented the Wasserstein loss function together with gradient penalty.
An application that webscrapes live data from the Robert Koch-Institut. It then displays the incidence rates or total cases for Germany as a map.
Stock analyzing program with live data coming from Yahoo finance. I also use fbprophet to predict upcoming prices.
Dueling-Double-Deep-Q-Network playing Space Invaders. The environment is built in PyGame. And the Agent keeps perpetually improving by observing the game.
Within the 10 days of a school project, I built an interactive website on the topic of renewable energy sources. Energie21
A simple Pygame application that helps you understand and visualize simple sorting algorithms like BubbleSort and InsertionSort.
Social Network
- Python
- Django
I built a Social Network with Django which connects to a backend and lets you create your own account and share posts. The hosting was done by Heroku.
Processing diary with own projects like 3d online multiplayer games and tasks from Michael Kipp.