Online LDA
Block Catalogue / Analytics
Streaming Linear Discriminant Analysis — maintains per-class means and within/between-class scatter online, projecting into a k-dim subspace that maximises class separation. Feeds the labelled scatter-plot UI.
Input and output
| Property | Value |
|---|---|
| Type | onlinelda, mosaic.models.onlinelda |
| Inputs | 2; see usage below |
| Consumes | Vector<double> or Matrix<double> (row-wise) on the data edge; Vector<double> (start) or null (stop) on the label edge |
| Publishes | ValueTuple<string, Vector<double>> — (label, k-dim projection) |
Parameters
| # | Name | Type | Default | Meaning |
|---|---|---|---|---|
| 0 | k | int | 2 | Components; must be 2 or 3 |
| 1 | eta0 | double | 0.1 | Learning rate |
| 2 | reorthEvery | int | 100 | Re-orthonormalisation interval |
| 3 | minStableCount | int | 500 | Samples before IsStable flips |
| 4 | regularization | double | 1e-4 | Added to the Sw diagonal |
Requirements and use
Provide labelled observations and enough classes/data before interpreting the projection. Keep the label path distinct from the numerical feature dimensions.
For parameter positions and connection rules, see Pipeline Configuration. For capturing output, see Recording.
Implementation
API reference. Model defaults above describe configuration loading; palette hints may expose a subset or select different defaults.