Visualization
We also offer a ml4chem.visualization
module to plot interesting
graphics about your model, features, or even monitor the progress of the loss
function and error minimization.
Two backends are supported to plot in ML4Chem: Seaborn and Plotly.
An example is shown below:
from ml4chem.visualization import plot_atomic_features
fig = plot_atomic_features("latent_space.db",
method="pca",
dimensions=3,
backend="plotly")
fig.write_html("latent_example.html")
This will produce an interactive plot with plotly where dimensionality was reduced using PCA, and an html with the name latent_example.html is created.
To activate plotly in Jupyter or JupyterLab follow the instructions shown in https://plot.ly/python/getting-started/#jupyter-notebook-support
If plotly is not rendering correctly you need to install the jupyter extension:
jupyter labextension install @jupyterlab/plotly-extension