Spaces:

nhamid289
/

decision_boundary-dev

Running

App Files Files Community

decision_boundary-dev / frontends /react /src /usage.md

nhamid289

Deploy decision_boundary from optimization-design-system (plain background, shared UI updates)

6c17bbb verified 4 days ago

preview code

Raw History Blame Contribute Delete

1.7 kB

### Draw your own data

1. Under **Data**, choose **Draw**.

2. Pick a class colour, then click on the plot to add points. Use a different colour for each class. **Undo** removes the last point and **Clear** removes them all.

3. Under **Model**, choose a model and adjust its settings. Each setting says what it does.

4. Press **Get Decision Boundary** to colour the plot by the class the model predicts. **Fit automatically** is on by default, so the model refits whenever the data or settings change; untick it to fit only when you press the button.

### Use a built-in or uploaded dataset

1. Under **Data**, choose **Built-in** and pick a dataset, or **Upload** a CSV file whose first row holds the column names.

2. Choose each column's role: **Input** (the model learns from it), **Target** (the class to predict) or **Off**. **View full table** shows every row.

3. Optionally normalise the inputs or add noise under **Preprocessing**.

4. With more than two inputs, choose what the plot shows under **Plot axes**: two of the inputs, or **PCA** (the two directions that keep the most variance).

5. Choose a model and press **Get Decision Boundary**, as above.

### Which data does the model see?

For a **drawn** dataset, the model is trained on the points on the plot.

For a **built-in or uploaded** dataset, the model is trained on all the input columns, after normalising and noise. The plot can only show two directions, so to colour it, each point on the plot is mapped back to the full inputs: with **Choose inputs**, the inputs that aren't shown are held at their mean; with **PCA**, the point is mapped back through the principal components. The caption under the plot says which.

	### Draw your own data

	1. Under Data, choose Draw.

	2. Pick a class colour, then click on the plot to add points. Use a different colour for each class. Undo removes the last point and Clear removes them all.

	3. Under Model, choose a model and adjust its settings. Each setting says what it does.

	4. Press Get Decision Boundary to colour the plot by the class the model predicts. Fit automatically is on by default, so the model refits whenever the data or settings change; untick it to fit only when you press the button.


	### Use a built-in or uploaded dataset

	1. Under Data, choose Built-in and pick a dataset, or Upload a CSV file whose first row holds the column names.

	2. Choose each column's role: Input (the model learns from it), Target (the class to predict) or Off. View full table shows every row.

	3. Optionally normalise the inputs or add noise under Preprocessing.

	4. With more than two inputs, choose what the plot shows under Plot axes: two of the inputs, or PCA (the two directions that keep the most variance).

	5. Choose a model and press Get Decision Boundary, as above.


	### Which data does the model see?

	For a drawn dataset, the model is trained on the points on the plot.

	For a built-in or uploaded dataset, the model is trained on all the input columns, after normalising and noise. The plot can only show two directions, so to colour it, each point on the plot is mapped back to the full inputs: with Choose inputs, the inputs that aren't shown are held at their mean; with PCA, the point is mapped back through the principal components. The caption under the plot says which.