A rapid introduction to Deepbox's core features — tensors, DataFrames, and machine learning in under 50 lines.

Learn the fundamentals of creating and manipulating tensors (N-dimensional arrays). Tensors are the core data structure in Deepbox.

Explore arithmetic, mathematical, and reduction operations on tensors. Deepbox supports 90+ tensor operations with full broadcasting.

Comprehensive data analysis workflow using DataFrames, statistics, and plotting. Explores, analyzes, and visualizes employee data.

Learn fundamental DataFrame operations for working with tabular data. Covers creation, selection, filtering, and sorting — for tabular data analysis.

Learn to group and aggregate data for analysis. Similar to SQL GROUP BY operations.

End-to-end ML pipeline demonstrating classification with the Iris dataset and regression with the Housing-Mini dataset. Includes data preprocessing, model comparison, cross-validation, and visualization.

Build a simple linear regression model to predict continuous values. Learn the basics of supervised learning with Deepbox.

Build a binary classification model using logistic regression. Learn to classify data into two categories using the Iris dataset.

Compare L1 (Lasso) and L2 (Ridge) regularization techniques. Learn when to use each regularization method.

Demonstrates advanced ML models: KMeans clustering, K-Nearest Neighbors (classification and regression), PCA dimensionality reduction, and Gaussian Naive Bayes.

Decision Trees, Random Forests, Gradient Boosting, and Linear SVM. Covers both classification and regression variants.

Bring everything together in a comprehensive machine learning workflow. From data loading to model evaluation and visualization.

Build and train neural networks using the nn, optim, and autograd modules. Covers Sequential models, custom modules, loss functions, and optimizers.

Deepbox's autograd system tracks operations on GradTensors to build a computation graph, then computes gradients via reverse-mode differentiation.

Explore different activation functions used in neural networks. Learn when to use each activation function.

Control the learning rate during training for better convergence. Deepbox provides 8 learning rate schedulers.

Transform categorical and label data into numeric representations using Deepbox's encoding utilities.

Feature scaling is essential before many ML algorithms. Deepbox provides 7 feature scalers.

Perform descriptive statistics and hypothesis testing. Analyze data distributions and test hypotheses.

Explore matrix decompositions and linear system solving. Essential for understanding ML algorithms under the hood.

Generate random numbers from various probability distributions. Useful for simulations, Monte Carlo methods, and data generation.

Explore Deepbox's 24 built-in datasets and 6 synthetic generators for quick experimentation. Perfect for learning and testing ML algorithms.

Learn different cross-validation techniques for robust model evaluation. Essential for assessing model generalization.

Learn to evaluate models using various performance metrics. Different metrics for classification, regression, and clustering.

Create various types of plots to visualize data and results. Deepbox supports SVG and PNG output for publication-quality figures.

Demonstrates CSR (Compressed Sparse Row) matrix operations. Memory-efficient representation for matrices with many zeros.

Demonstrates convolutional neural network layers: Conv1d, Conv2d, MaxPool2d, and AvgPool2d.

Demonstrates RNN, LSTM, and GRU layers for sequence modeling tasks.

Demonstrates MultiheadAttention and TransformerEncoderLayer for sequence-to-sequence modeling.

Demonstrates BatchNorm1d, LayerNorm, and Dropout for training stability and regularization.

Demonstrates the DataLoader class for efficient batch iteration over datasets.

Demonstrates the Module base class: custom modules, parameter registration, state serialization, train/eval modes, freeze/unfreeze, and Sequential container.

Advanced DataFrame operations new in v1.0.0: string/datetime accessors, rolling/expanding/EWM windows, query/eval expressions, pivot tables, crosstabs, and more.

Advanced ensemble methods and ML models new in v1.0.0: AdaBoost, Bagging, Voting, Stacking, ExtraTrees, Gaussian Processes, Discriminant Analysis, and Semi-supervised Learning.

Advanced clustering algorithms new in v1.0.0: Agglomerative, GaussianMixture, SpectralClustering, OPTICS, MiniBatchKMeans, MeanShift, Birch, and AffinityPropagation.

Support Vector Machines with kernel tricks and anomaly detection algorithms new in v1.0.0.

Automated model selection and ML pipelines new in v1.0.0: GridSearchCV, RandomizedSearchCV, Pipeline, ColumnTransformer, and cross-validation.

Advanced preprocessing tools new in v1.0.0: imputation, feature selection, text vectorizers, spline transformers, and advanced splitters.

Advanced NN features new in v1.0.0: Trainer with EarlyStopping, weight initialization, Embedding layers, normalization layers, containers, and advanced activations.

Full Transformer implementation new in v1.0.0: MultiheadAttention, TransformerEncoder/Decoder, PositionalEncoding, and complete encoder-decoder models.

Advanced optimizers and learning rate schedulers new in v1.0.0: RAdam, LAMB, LARS, CyclicLR, CosineAnnealingWarmRestarts, PolynomialLR, LambdaLR, and SequentialLR.

FFT and signal processing tools new in v1.0.0: fft, ifft, rfft, irfft, fft2, ifft2, fftn, and spectral analysis utilities.

Comprehensive statistics module new in v1.0.0: probability distributions, hypothesis tests, correlations, and confidence intervals.

Visualization tools new in v1.0.0: line plots, scatter, histograms, heatmaps, confusion matrices, ROC curves, learning curves, decision boundaries, dendrograms, and 3D scatter.

A runtime-focused walkthrough for the v1.0.0 `deepbox/core` surface: structured logging, warning policies, backend registration, and JSON/file serialization.

A data-pipeline example for the v1.0.0 dataset helpers around training loops: `Subset`, `randomSplit`, `mapDataset`, `filterDataset`, `WeightedRandomSampler`, and `SubsetRandomSampler`.

A practical DataFrame operations example for the v1.0.0 polish layer: JSON/XLSX/Parquet round-trips, datetime parsing, styling, and pandas-like plotting accessors.

A focused walkthrough for the v1.0.0 inference APIs that were still missing from the runnable docs: confidence intervals, bootstrap resampling, Gaussian KDE, multiple-comparison correction, and statistical power planning.

A focused v1.0.0 linear algebra example covering the new advanced routines missing from the earlier decomposition walkthrough: Hessenberg and Schur decompositions, polar decomposition, matrix functions, structured solvers, sparse CSR solving, and Sylvester/Lyapunov equations.