DEWS-IV
This algorithm is designed for datasets and models where consistency among neighbors is important. It is a variation of DEWS-V that takes neighbor distance into consideration. It uses soft blending between the top experts in a certain competence region to compute a set of weights for the models.
When to use
- DEWS-IV is only recommended for specific cases and not recommended as a general algorithm
- It performs best when a dataset has regions where models should be consistent throughout. It can perform well when pools are diverse with no dominant model and datasets are medium-sized with moderate dimensionality and have a structured feature space.
- It performs worst for small values of k, since few data points can make variability measures noisy. It also struggles when datasets don't have clear region boundaries, so a strong model may have variability.
How it works
When fit is called, DEWS-IV fits a KNN algorithm on the validation data and builds a criterion score matrix.
When predict is called, it finds the K nearest neighbors from the test point and uses the score matrix to combine
every models' scores over the K neighbors with inverse-distance weights and subtracts a penalty term proportional
to the variability from the scores.
Afterwards, it normalizes the average scores using min-max normalization and removes the models under a threshold.
Finally, it takes the remaining models and creates weights with their scores using softmax with temperature.
These weights can then be provided as output or combined with predictions to make a final ensembled answer.
Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
task |
str | — | "classification" or "regression" |
metric |
str or callable | — | Scoring function per sample. Built-ins: accuracy, mae, mse, rmse, log_loss, prob_correct. Custom callables (y_true, y_pred) -> float are accepted |
mode |
str | — | "max" if higher is better, "min" if lower |
k |
int | 10 | Number of neighbours |
threshold |
float | 0.5 | Competence cutoff |
temperature |
float | 0.5/1.0 for regression/classification | Defines how smooth the model blend is |
distance_metric |
str | "euclidean" | Distance metric used for KNN/ANN. See distance metrics |
preset |
str | "balanced" |
ANN backend preset. Options: "exact","balanced", "fast", "turbo", "high_dim_balanced", "high_dim_fast" |
finder |
str | —, optional | Only if the preset is "custom"; Options: "knn", "faiss", "annoy", "hnsw" |
loo |
bool | false | Enables Leave One Out (LOO) tuning; ignores the closest neighbor with negligible distance when selecting the K neighbors. |
Example
# Regression
from deskit.des.dewsiv import DEWSIV
router = DEWSIV(task="regression", metric="mae", mode="min", k=20)
router.fit(X_val, y_val, val_preds)
weights = router.predict(x)
# Classification
from deskit.des.dewsiv import DEWSIV
router = DEWSIV(task="classification", metric="log_loss", mode="min", k=20)
router.fit(X_val, y_val, val_preds)
answers = router.predict(X_test, test_preds)
Notes
A lower temperature is recommended for regression because regression metrics tend to produce scores on a continuous scale where differences can be large, so a low temperature sharpens the softmax to reflect that. In contrast, classification metrics tend to produce scores that are closer together, so a higher temperature keeps the blend soft.