Adapt Algo Selector¶

adapt_algo_selector — mvp.adapt_algo_selector

Cluster: Uncategorised | Type: component | MCP Tools: 13

Overview¶

Algorithm selector that extracts dataset meta-features (sample count, feature count, class imbalance, linearity) and applies rule-based heuristics to produce ranked algorithm recommendations for classification, regression, and clustering tasks. This is the canonical public name for the component backed by mvp.adapt_learning, re-exporting AdaptAlgoSelectorBlock, AlgoSelectorInput, AlgoSelectorOutput, and AlgorithmRecommendation. Pure Python with no external ML dependencies required.

When to use:

Selecting an appropriate ML algorithm for a given dataset without manual benchmarking
Getting ranked algorithm recommendations based on meta-feature analysis
Integrating algorithm selection into AutoML pipelines

Example:

from mvp.adapt_algo_selector import AdaptAlgoSelectorBlock, AlgoSelectorInput

block = AdaptAlgoSelectorBlock(name="algo_selector")
result = block.infer(AlgoSelectorInput(
    X=[[1.0, 2.0], [2.0, 3.0], [3.0, 4.0], [4.0, 5.0], [5.0, 6.0],
       [1.5, 2.5], [2.5, 3.5], [3.5, 4.5]],  # n_samples (rows) × n_features (cols)
    y=[0, 0, 0, 1, 1, 0, 1, 1],                # length n_samples; class labels for classification
    task="classification",                       # or "regression" / "clustering" / "auto"
))
# result.value.recommendations →
#   [KNeighborsClassifier (0.72), RandomForestClassifier (0.70), LogisticRegression (0.60)]
# result.value.dataset_properties → {n_samples: 8, n_features: 2, sample_feature_ratio: 4.0,
#                                    class_imbalance: 1.0, n_classes: 2, ...}

Required input shape: X is a list-of-lists of numeric features (list[list[float | int]]), shape n_samples × n_features. y is a list[float | int | str] of length n_samples. task accepts "classification", "regression", "clustering", or "auto" (default infers from y dtype).

Works well with: adapt_automl, adapt_sklearn, adapt_bayesian, adapt_pandas

Public API¶