georgetown-analytics / data-space Goto Github PK

Add multilayer @perceptron hyperparams and help text

Use yellowbrick to generate a classification report, confusion matrix, precision-recall curve, etc. and send them as base64 encoded data for display in the dataspace.

Have a spinner appear and disable the form during fit so multiple fit calls don't spam the server and confuse the user.

Add logistic regression hyperparams and help text

Add k nearest neighbors hyperparams and help text

Right now the server sends back a 400 with a message body that says what went wrong. However, D3.json cannot read the error body. We should send back a JSON response with error in it that d3 understands so we can have better alerts.

Update the Naive Bayes modal with help text.

A short tutorial. I imagine if it is the person's first visit, an automated tutorial goes through a step-by-step sequence of how to visualize the model

1. Generate Data - Select Binary,multiclass,etc
2. Select model type
3. Select specific model
4. Select parameters
5. Click fit
6. Wait

In the LogisticRegression form, when fit_intercept is False, disableintercept_scaling.

Right now we are fetching dependencies from the CDN - which is great; less hosting on our part! We should cache what we have locally though just in case for the future.

Add SVM form hyperparms and help text.

We need a favicon!

The original svm_gui.py used a contour map to display the decision space (at least for SVM). Potentially we could do something similar with D3:

https://observablehq.com/@d3/contours?collection=@d3/d3-contour