GPInter

Gaussian Processes from scratch with interactive visualization.

Install requirements

pip install -r requirements.txt

Usage

cd GPInter
python main.py

• Left click on the plot to add a new point (new points are added with 0 error)
• Right click on the plot to remove a point (deletes the nearest point to cursor)
• Use sliders to change length scale and amplitude (+ other kernel parameters if available)
• Reset button resets the plot to initial state
• Kernel radio button to choose different kernels

TODO

• Refactor code and implement Kernel base class
• Implement more kernels
• Drop down menu to choose different kernels on the fly
• (Re)Implement plot_gp for notebook
• Implement kernel operations (maybe at the cost of interactivity)
• Implement hyperparameter optimization (does not work with combined kernels)
• Add tests (does it need one?)