correlation_view

Back to L3.5 | Browse all axes | Browse all options

Axis correlation_view on sub-layer L3_5_C_feature_correlation (layer l3_5).

Sub-layer

L3_5_C_feature_correlation

Axis metadata

  • Default: 'clustered_heatmap'

  • Sweepable: False

  • Status: operational

Operational status summary

  • Operational: 3 option(s)

  • Future: 0 option(s)

Options

clustered_heatmap – operational

Clustered heatmap reordered by hierarchical clustering.

L3.5.C correlation view clustered_heatmap.

This option configures the correlation_view axis on the L3_5_C_feature_correlation sub-layer of L3.5; output is emitted under manifest.diagnostics/l3_5/L3_5_C_feature_correlation/ alongside the other selected views.

When to use

Large feature panels with block structure; reveals clusters of correlated features.

References

  • macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’

Related options: full_matrix, top_k

Last reviewed 2026-05-05 by macroforecast author.

full_matrix – operational

Full feature × feature correlation matrix.

L3.5.C correlation view full_matrix.

This option configures the correlation_view axis on the L3_5_C_feature_correlation sub-layer of L3.5; output is emitted under manifest.diagnostics/l3_5/L3_5_C_feature_correlation/ alongside the other selected views.

When to use

Small feature panels (< 100 cols).

References

  • macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’

Related options: clustered_heatmap, top_k

Last reviewed 2026-05-05 by macroforecast author.

top_k – operational

Top-k highest-|ρ| pairs.

L3.5.C correlation view top_k.

This option configures the correlation_view axis on the L3_5_C_feature_correlation sub-layer of L3.5; output is emitted under manifest.diagnostics/l3_5/L3_5_C_feature_correlation/ alongside the other selected views.

When to use

Cheapest readout for very wide panels.

References

  • macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’

Related options: full_matrix, clustered_heatmap

Last reviewed 2026-05-05 by macroforecast author.