🤖

AI & Machine Learning

Neural networks, evolutionary algorithms, reinforcement learning and classification — explore the core ideas of artificial intelligence, visualised.

This category covers the foundations of artificial intelligence and machine learning: how computers learn patterns from data, optimise their own parameters, make decisions, and improve through experience. Each interactive AI & Machine Learning model runs live in your browser, so you can adjust hyperparameters, swap datasets and watch backpropagation, Q-learning, genetic algorithms, clustering and decision trees unfold step by step. You will build genuine intuition for supervised, unsupervised and reinforcement learning, understand why models overfit, and see how gradients drive training. Whether you are a student, developer or curious learner, these simulations help you learn AI & Machine Learning online — the same techniques that power recommendation engines, computer vision, fraud detection and large language models in the real world.

8 simulations Canvas 2D · WebGL Uses Backpropagation, Q-Learning, CART

🤖 See also: Machine Learning Topic Hub

AI & ML Simulations

Open a simulation — it runs right in your browser, no installation needed

🌳
Popular ★☆☆ Beginner
Decision Tree
Live CART decision tree visualiser. Draw data points, watch the tree grow with Gini or entropy splits, and see the classification boundary update in real time.
Canvas 2D CART Classification
🧠
★★☆ Moderate
Neural Network
Interactive backpropagation visualiser. Build a multi-layer perceptron, pick a dataset, train the network and watch weights, biases, and the decision boundary evolve in real time.
Canvas 2D Backprop Perceptron
🧬
★★☆ Moderate
Genetic Algorithm
Evolve solutions through selection, crossover, and mutation. Fitness-proportional reproduction optimises a target string or function — watch the population converge over generations.
Canvas 2D Evolution Optimisation
🎮
New ★★☆ Moderate
Reinforcement Learning
Q-learning agent navigates a grid world with rewards and traps. Watch the Q-table update and the optimal policy emerge through exploration-exploitation over episodes.
Canvas 2D Q-Learning Grid World
🗺️
★★☆ Moderate
Self-Organising Map
Kohonen SOM learns the topology of colour space. A 2D grid of neurons adapts to 3D RGB input, producing a smooth colour map that preserves neighbourhood relationships.
Canvas 2D Kohonen Unsupervised
🦎
★☆☆ Beginner
Natural Selection
Creatures with heritable traits compete for limited resources. Speed, size, and camouflage are subject to selection pressure — watch adaptation and speciation unfold.
Canvas 2D Evolution Selection
🧭
★☆☆ Beginner
Pathfinding
A*, Dijkstra, BFS and DFS on a weighted grid. Draw walls and weights, place start and goal, then watch the algorithm explore and find the shortest path.
Canvas 2D A* Graph Search
📉
New ★★☆ Intermediate
Gradient Descent Visualiser
Watch SGD, Momentum, RMSprop and Adam navigate a 3D loss landscape in real time. Compare convergence speed and trajectory shape across four optimizers on Rosenbrock, Saddle, Beale and Himmelblau functions.
Canvas 2D Optimization Machine Learning
🤖
New ★★☆ Intermediate
K-Means Clustering
Watch K-Means partition data into Voronoi regions step by step. Use K-Means++ initialisation, run the elbow method to find optimal K, click to add custom points, or generate blobs, half-moons and ring datasets.
K-Means++ Voronoi Unsupervised Learning
🧠
New ★★☆ Intermediate
Convolutional Neural Network
Visualise CNN layers step by step — convolution filters, ReLU activations and max-pooling. Watch feature maps emerge as a filter scans an 8×8 image.
CNN Deep Learning Feature Maps
🧠
New ★★★ Advanced
Neural Gas
Watch a neural gas network adapt to data topology — competitive learning with neighbourhood ranking. Nodes drift to represent the underlying distribution.
Neural Gas Unsupervised Topology
🤖
New ★★☆ Intermediate
Neural Network Training
Watch backpropagation train a small feedforward network in real time — see weights and loss evolve epoch by epoch with adjustable learning rate and architecture.
Backpropagation Deep Learning Training

About AI & Machine Learning Simulations

Learn, evolve, decide — the algorithms that power intelligent systems

AI and machine learning simulations make the inner workings of intelligent algorithms visible and interactive. Rather than treating neural networks, decision trees, and reinforcement learning agents as black boxes, these visualisations show exactly how each algorithm processes data, updates its parameters, and improves its performance step by step.

The genetic algorithm demonstrates evolution in silico — populations of candidate solutions undergo selection, crossover, and mutation to solve optimisation problems. Self-organising maps reveal how an unsupervised network learns the topology of high-dimensional data by projecting it onto a 2D grid, while reinforcement learning shows an agent discovering optimal policies through trial-and-error in a grid world.

These are the same core techniques powering modern AI: backpropagation trains deep learning models, Q-learning underpins game-playing agents, and decision trees remain a go-to for interpretable classification. Running them in a browser lets you experiment with hyperparameters, datasets, and architectures to build genuine intuition about what each algorithm can — and cannot — learn.

Key Concepts

Topics and algorithms you'll explore in this category

BackpropagationComputing gradients through a neural network via the chain rule
Q-LearningModel-free reinforcement learning via temporal difference updates
Genetic AlgorithmSelection, crossover, and mutation for evolutionary optimisation
Decision TreeRecursive feature splitting for interpretable classification
Self-Organising MapUnsupervised topology-preserving dimensionality reduction
Natural SelectionFitness-proportional reproduction with heritable variation

🤖 Test Your AI & ML Knowledge

Five quick questions to check your understanding of artificial intelligence and machine learning

AI & ML Quiz

Frequently Asked Questions

Common questions about this simulation category

How does backpropagation work?
Backpropagation computes the gradient of the loss with respect to each weight by applying the chain rule layer by layer, from output back to input. Each weight is then adjusted proportionally to its gradient, reducing the network's error on training examples. The visualiser shows how the decision boundary shifts with each epoch.
What is Q-learning?
Q-learning is a model-free reinforcement learning algorithm. An agent explores a grid world, receiving rewards and penalties. It maintains a Q-table that estimates the expected future reward for each state-action pair, updating it via the Bellman equation after each step. Over time, the Q-values converge to the optimal policy.
How do genetic algorithms find solutions?
A population of random candidate solutions is evaluated by a fitness function. The fittest individuals are selected for reproduction, recombined via crossover, and randomly mutated. Over generations, the population converges toward high-fitness solutions — mimicking biological evolution.

Other Categories

Every AI & Machine Learning simulation in this collection turns abstract theory into something you can see and control. Each interactive AI & Machine Learning model lets you draw data, tune parameters and observe how an algorithm converges in real time — making it far easier to learn AI & Machine Learning online than from equations alone. These are the very methods behind real-world applications such as medical image diagnosis, spam filtering, self-driving navigation and product recommendations. Run them directly in your browser, with no installation or sign-up, to build lasting intuition about how intelligent systems train, generalise and make predictions from data.