Visual lab for teaching

Interactive platform.

Choose an experiment, open fullscreen mode, enter your own data, or use automatic animation to explain concepts step by step.

ML Methods

Core algorithms from dimensionality reduction to deep learning.

PCADimensionality ReductionStarter

PCA

Choose 2 or 3 principal components and compare PCA projections in 2D and 3D using explained variance.

Z = XW
KNNClassificationPractitioner

KNN

Add points, change K, and observe how nearest neighbors shape the decision boundary.

d(x,y) = sqrt(sum((x-y)^2))
KMClusteringPractitioner

K-Means

Run automatic or step-by-step iterations to follow centroid movement.

min sum ||x - mu_k||^2
GDOptimizationStarter

Gradient Descent

Pick a starting point, compare learning rates, and follow the path toward local or global minima.

theta <- theta - alpha * grad L
LRRegressionStarter

Linear Regression

Enter your own observations and watch the best-fit line and residuals update.

y = beta_0 + beta_1 x
LOGClassificationStarter

Logistic Regression

Learn how the sigmoid function converts a linear score into class probability.

p(y=1) = 1 / (1 + exp(-z))
DTClassificationFree Preview

Decision Tree

Prepare a dataset, run training, then compare data boundaries with tree rules and Gini calculations.

Gain = Gini(parent) - sum w_i Gini(child_i)
RFEnsemblePractitioner

Random Forest

Compare voting across many trees and see how an ensemble boundary becomes more stable.

y = mode(h_1(x), ..., h_T(x))
SVMClassificationPractitioner

SVM

Visualize the hyperplane, margin, and support vectors that define the decision boundary.

min 1/2 ||w||^2 + C sum hinge loss
NNDeep LearningComplete

Neural Network

Configure hidden layers, train an XOR network, and inspect neuron activations inside the architecture.

a = sigma(Wx + b)
CNNDeep LearningComplete

CNN Convolution

Edit input pixels, choose a kernel, and watch feature maps produced by convolution.

Y[i,j] = sum_m sum_n X[i+m,j+n] K[m,n] + b
ATTNDeep LearningComplete

Transformer

Enter your own sentence and inspect Q, K, V, scaled dot-product, softmax, and the attention matrix.

Attention(Q,K,V) = softmax(QK^T / sqrt(d))V