Plot Neural Network Classification Landscapes

Plots landscapes with neural network classification results, highlighting correct and misclassified cases. Optionally, only misclassified landscapes can be shown.

Usage

plot_classified_landscapes(
  classification,
  landscapes,
  only_misclassified = FALSE,
  score_note = TRUE,
  subset_index = NULL,
  ...
)

Arguments

classification

A data frame with columns: landscape_id, actual_class, predicted_class, and score. Can be obtained from the CV-fold results of train_metric_model/train_pixel_model or the output of apply_metric_model/apply_pixel_model.

landscapes

A list of landscape objects corresponding one-to-one and in the same order as the rows in `classification`. The easiest way to ensure this is to use the same list of landscapes for both training and plotting.

only_misclassified

Logical; if TRUE, only misclassified landscapes are plotted. Default is FALSE. If every landscape was classified correctly there is nothing to plot: the function reports this with a message and returns an empty placeholder plot. Landscapes whose true class is unknown are not counted as misclassified.

score_note

Logical; if TRUE (default), a one-line caption is added under the whole figure stating that the bracketed number is the score of the predicted class and not a calibrated probability. Set to FALSE when the surrounding figure caption already says so.

subset_index

Integer vector. Which of the plotted landscapes to show, e.g. to keep a large figure readable. Indexes the rows of classification that would otherwise be plotted, so with only_misclassified = TRUE it selects among the misclassified ones. Default NULL plots all of them.

...

Additional arguments passed to plot_landscapes, such as show_legend, legend_title, ncol, max_landscapes, or force.

Value

A patchwork object combining landscape plots with classification annotations. With only_misclassified = TRUE and no misclassified landscape, an empty placeholder plot carrying that message.

See also

train_pixel_model, train_metric_model

Other visualization: plot_landscapes(), plot_metrics()

Examples

# \donttest{
# Generate training landscapes
landscapes <- create_landscapes(
  n = 18,
  patterns = c("random", "sharp", "diffuse")
)
#> ✔ Successfully generated all 18 training landscapes

# Calculate landscape metrics
metrics <- calculate_metrics(landscapes, level = "landscape")
#>  ■■■■■■■■■■                        30% |  ETA:  7s
#>  ■■■■■■■■■■■■■■■                   47% |  ETA:  7s
#>  ■■■■■■■■■■■■■■■■■■■               59% |  ETA:  6s
#>  ■■■■■■■■■■■■■■■■■■■■■■■■■■■       88% |  ETA:  2s

# Find the best 5 metrics for classification
best_5 <- evaluate_metrics(metrics, metrics_number = 5)
#> Warning: Excluded 6 metrics with missing values (108 rows removed).
#> ✖ NA value for at least one landscape: "enn_cv", "enn_mn", "enn_sd", "iji",
#>   "pafrac", and "rpr"
#> ℹ Use `exclude_incomplete_metrics = FALSE` to retain them (not recommended for
#>   model training).
#> Warning: Excluded 3 metrics with no variation across landscapes: "pr", "prd", and "ta"

# Cross-validation produces the held-out predictions this plot needs.
# Only 2 folds, as each fold needs at least 3 landscapes per pattern.
model <- train_metric_model(
  metrics,
  metrics_selected = best_5,
  cv_method = "k-fold",
  cv_folds = 2
)
#> ℹ Low sample-to-predictor ratio (3.6:1). Consider LOO CV or reducing features.
#> 
#> ── Cross-validation results ──
#> 
#> ℹ Method: 2-fold cross-validation
#> ℹ Overall accuracy: 94.44%
#> 
#> ── Confusion matrix 
#>          Actual
#> Predicted diffuse random sharp
#>   diffuse       6      0     0
#>   random        0      5     0
#>   sharp         0      1     6
#> 
#> ── Per-class performance 
#> # A tibble: 3 × 5
#>   class   count recall precision f1_score
#>   <chr>   <dbl>  <dbl>     <dbl>    <dbl>
#> 1 diffuse     6   1         1        1   
#> 2 random      6   0.83      1        0.91
#> 3 sharp       6   1         0.86     0.92

# Plot all classification results
plot_classified_landscapes(
  model$performance$validation_results,
  landscapes
)


# Show only misclassifications without legend
plot_classified_landscapes(
  model$performance$validation_results,
  landscapes,
  only_misclassified = TRUE,
  show_legend = FALSE,
  ncol = 4
)

# }