Adds two notebooks to a folder examples/by-hand that allow users to 1) fit a linear model by hand and 2) fit a local landmarks model by hand. Local landmarks is a cousin of nearest-neighbors that is not restricted to using points from the dataset. The latter demonstrates over-fitting.
The "by-hand" aspect is implemented with interactive matplotlib figures controlled by ipywidgets. The heavy lifting is done in utils/models.py and utils/landmarks.py.
Support for these was recently added to the hub, but isn't currently guaranteed by the requirements.txt of this repo. I can add what's needed, but currently requirements doesn't even include Jupyter, and seems aimed at supporting the projects.
Both notebooks include wandb tracking, with the settings pointing to the qualcomm entity, as in the other recently-used example sub-dirs. This may cause an error until an appropriate public project is created.
Adds two notebooks to a folder
examples/by-handthat allow users to 1) fit a linear model by hand and 2) fit a local landmarks model by hand. Local landmarks is a cousin of nearest-neighbors that is not restricted to using points from the dataset. The latter demonstrates over-fitting.The "by-hand" aspect is implemented with interactive matplotlib figures controlled by ipywidgets. The heavy lifting is done in
utils/models.pyandutils/landmarks.py.Support for these was recently added to the hub, but isn't currently guaranteed by the
requirements.txtof this repo. I can add what's needed, but currently requirements doesn't even include Jupyter, and seems aimed at supporting theprojects.Both notebooks include wandb tracking, with the
settingspointing to thequalcommentity, as in the other recently-used example sub-dirs. This may cause an error until an appropriate public project is created.