Classical Machine Learning
155+The most comprehensive classical ML toolkit available for .NET
Everything from random forests to XGBoost, from K-Means to HDBSCAN, from isolation forests to deep anomaly detectors. AiDotNet provides more classical ML algorithms than scikit-learn, all in pure C# with full GPU acceleration and SIMD optimization support.
Classification (52 algorithms)
Predict categorical outcomes with ensemble methods, SVMs, Bayesian classifiers, and more.
RandomForest
Bagged ensemble of decision trees with feature importance and OOB estimation.
GradientBoosting
Sequential boosting with gradient descent optimization.
XGBoost
Extreme gradient boosting with regularization and histogram-based splitting.
LightGBM
Gradient boosting with leaf-wise growth and categorical feature support.
CatBoost
Gradient boosting with native categorical feature handling.
SVM
Support vector machine with linear, RBF, polynomial, and sigmoid kernels.
KNN
K-nearest neighbors with distance weighting and ball/KD tree acceleration.
LogisticRegression
Linear classification with L1/L2 regularization and multinomial support.
NaiveBayes
Gaussian, Multinomial, and Bernoulli Naive Bayes classifiers.
ExtraTrees
Extremely randomized trees for faster training with comparable accuracy.
AdaBoost
Adaptive boosting with automatic re-weighting of misclassified samples.
StackingClassifier
Meta-learning ensemble combining multiple base classifiers.
Regression (59 algorithms)
Predict continuous values with linear models, tree ensembles, kernel methods, and robust estimators.
LinearRegression
Ordinary least squares with closed-form solution and statistical diagnostics.
Ridge
L2-regularized regression for handling multicollinearity.
Lasso
L1-regularized regression with automatic feature selection.
ElasticNet
Combined L1+L2 regularization with adjustable mixing ratio.
RandomForestRegressor
Ensemble of regression trees with variance reduction.
GradientBoostingRegressor
Sequential boosting for regression with multiple loss functions.
SVR
Support Vector Regression with epsilon-insensitive loss.
BayesianRidge
Bayesian linear regression with automatic relevance determination.
HuberRegressor
Robust regression resistant to outliers using Huber loss.
QuantileRegression
Predict conditional quantiles for uncertainty estimation.
IsotonicRegression
Non-parametric monotonic regression for calibration.
Clustering (90+ algorithms)
Discover natural groupings with density-based, centroid-based, hierarchical, and spectral methods.
KMeans
Centroid-based clustering with k-means++ initialization.
DBSCAN
Density-based clustering that discovers arbitrary-shaped clusters.
HDBSCAN
Hierarchical DBSCAN with automatic cluster selection and stability scores.
Agglomerative
Bottom-up hierarchical clustering with single, complete, average, Ward linkage.
Spectral
Graph-based clustering using eigenvalues of similarity matrix.
GaussianMixture
Probabilistic clustering with EM algorithm and model selection (BIC/AIC).
OPTICS
Ordering Points to Identify Clustering Structure for varying density.
MeanShift
Non-parametric clustering that finds modes in density distribution.
Birch
Balanced Iterative Reducing and Clustering using Hierarchies for large datasets.
AffinityPropagation
Message-passing algorithm that simultaneously identifies exemplars and clusters.
MiniBatchKMeans
Scalable K-Means variant using mini-batches for large datasets.
Anomaly Detection (66 algorithms)
Identify outliers and anomalies with isolation-based, density-based, and deep learning approaches.
IsolationForest
Anomaly detection via random partitioning of feature space.
LocalOutlierFactor
Density-based anomaly detection comparing local reachability densities.
OneClassSVM
Support vector method for novelty detection in high-dimensional spaces.
EllipticEnvelope
Gaussian assumption-based outlier detection using Mahalanobis distance.
AutoEncoder
Neural network-based anomaly detection via reconstruction error.
COPOD
Copula-based outlier detection - fast, parameter-free.
ECOD
Empirical cumulative distribution-based outlier detection.
DeepSVDD
Deep Support Vector Data Description for complex data.
DAGMM
Deep Autoencoding Gaussian Mixture Model for unsupervised anomaly detection.
TranAD
Transformer-based adversarial anomaly detection for time series.
Dimensionality Reduction
Reduce feature space while preserving structure for visualization and preprocessing.
PCA
Principal Component Analysis for linear dimensionality reduction.
KernelPCA
Non-linear PCA using kernel trick for complex manifolds.
t-SNE
T-distributed Stochastic Neighbor Embedding for 2D/3D visualization.
UMAP
Uniform Manifold Approximation for fast, high-quality embeddings.
LDA
Linear Discriminant Analysis for supervised dimensionality reduction.
NMF
Non-negative Matrix Factorization for parts-based decomposition.
ICA
Independent Component Analysis for blind source separation.
Isomap
Isometric mapping preserving geodesic distances on manifolds.
Classification with AiModelBuilder
using AiDotNet;
// Classification with AiModelBuilder
var result = await new AiModelBuilder<float, float[], float>()
.ConfigureModel(new GradientBoostingClassifier<float>(
numTrees: 200, maxDepth: 6, learningRate: 0.1f))
.ConfigurePreprocessing()
.BuildAsync(X_train, y_train);
var predictions = result.Predict(X_test);Start building with Classical Machine Learning
All 155+ implementations are included free under Apache 2.0.