Machine Learning

Synthetic patient cohorts, virtual clinical trials of medical algorithms and an AI safety audit.

Convolutional Neural Networks: slide 1D and 2D kernels over signals and images, step by step, with activation and pooling.

Class-based data augmentation with CSV or sample dataset: SMOTE, Borderline-SMOTE, ADASYN, mixup, jitter, Gaussian copula and browser-trained small GAN; The result is downloaded as a file labeled original/synthetic.

Interactive laboratory for detecting data drift and its effects.

Interactive illustration of sources of data leakage and impact on model performance.

Step-by-step interactive simulator of the ML data preprocessing pipeline.

Interactive visualization of decision boundaries of different classifiers.

Examining early stopping point via training/validation error.

Comparison of hold-out, k-fold, LOO and bootstrap validation methods.

Interactive laboratory of grid, random and Bayesian hyperparameter optimization methods.

k-nearest neighbor classification with your own CSV data or sample datasets: enter k, distance measure, weighting, and scaling parameters or use defaults; Examine results with metrics, confusion matrix, permutation significance, and k-scan plot.

Interactive laboratory that explains what logistic regression is (sigmoid, log-odds, log loss, L2 regularization, decision threshold) and trains the model with your own CSV data or sample data sets and examines the results with metrics, confusion matrix, coefficient/odds ratios, ROC curve and loss plot.

Interactive MLOps lab of machine learning workflow (data, training, validation, deployment).

Audit a clinical ML study: data checks for leakage and shortcuts plus an AI peer-review-style safety report.

ML pipeline and model selection simulator.

Interactive neural network training environment based on TensorFlow.js.

Precision, recall, F1, MCC, AUC-ROC and balanced accuracy calculation and learning tool for binary classification.

Visualization of resampling methods such as hold-out, k-fold, bootstrap.

Effect of clinical decision threshold on sensitivity/specificity.

Comparison of ROC and Precision-Recall curves on unbalanced data sets.

Upload SHAP plots and tables for clear, plain-English interpretations of the results.

Analysis tool that automatically selects the statistical test appropriate to the data (t-test, Mann–Whitney, Wilcoxon, ANOVA, Kruskal–Wallis, chi-square, correlation), explains why, and calculates p-value, effect size and confidence intervals.

Interactive learning module of TabPFN transformer based classifier for tabular data.

Explore how model complexity affects underfitting and overfitting.

Gradient boosting classification with your own CSV data or sample datasets: enter parameters such as n_estimators, learning rate, max_depth, gamma, lambda/alpha, subsample, or use the defaults; Examine results with metrics, confusion matrix, and attribute importance.