Advanced Concepts of Modeling in AI
How machines learn: from rules to neural networks. Play with each visual, then start its quiz.
1AI, ML and DL
Artificial Intelligence is the big idea of machines acting smart. Machine Learning is a part of AI where machines learn from data. Deep Learning is a part of ML that uses many-layered neural networks.
Place each example. Tap an example, then tap the circle where it belongs.
2Rule-based vs learning-based models
In a rule-based model, a human writes the rules. In a learning-based model, the machine finds the rule from labelled examples, and can learn again when the data changes. Sort apples from oranges both ways.
Rule builder. A fruit is an apple if it matches any rule below, otherwise it is an orange.
The model starts with a random line. It looks at one fruit at a time. If it is wrong, it nudges the line towards that fruit.
Accuracy after each round (1 round = every fruit checked once)
Rule it found:
3Supervised, unsupervised and reinforcement learning
Machines learn in three main ways: from labelled examples, by finding patterns in unlabelled data, or by trial, error and reward.
Every dot has a label: ● Cat or ▲ Dog (by ear length and weight). The model learns the coloured regions from these labels.
Tap anywhere on the plot to add a new animal. The model predicts its label.
No labels this time. The machine groups similar points on its own (clustering with k-means).
The robot 🤖 must reach the star ⭐ and avoid the pits. It gets +10 for the star, −10 for a pit and −1 per step. Nobody tells it the path; it learns from rewards.
Games played: 0 · Last reward: –
4Classification, regression, clustering and association
Supervised learning splits into classification (predict a category) and regression (predict a number). Unsupervised learning splits into clustering (find groups) and association (find items that go together).
Will a student pass? Each dot is a past student, labelled ● Pass or ✖ Fail. The model learns a boundary from these labels.
Training steps: 0 · Accuracy:
Predict the marks. Fit a straight line through the data. Each orange square shows one error squared; smaller total = better line.
Tip: drag any dot to change the data.
Make T-shirt sizes for the school. Nobody has labelled these students. k-means finds groups of similar height and weight, and each group becomes a size.
Total distance to centres (lower = tighter groups):
School canteen, 20 bills. Which items are bought together? Tap a square in the grid to test a rule, or build your own tray below.
5How does AI make a decision?
One artificial neuron multiplies each input by a weight (how important it is), adds them up with a bias, and decides. This is the building block of every neural network.
Should I buy this phone? Turn inputs on or off, and change how much each one matters to you.
6Artificial Neural Network, layer by layer
Neurons are arranged in layers. The input layer takes the data, hidden layers find patterns, and the output layer gives the answer. Information flows forward, one layer at a time.
Lines: blue = positive weight, orange = negative weight, thicker = stronger. Brighter neuron = more active. Tap a neuron to see its maths.
7Convolutional Neural Network (CNN)
A CNN is a neural network for images. A small grid of numbers called a kernel slides over the picture to find features like edges. Pooling then shrinks the result, and the final layers decide what the picture shows.
Try it yourself
Tools named in the CBSE syllabus for this unit: Teachable Machine (supervised), Google Drum Machine (unsupervised), TensorFlow Playground (neural networks). Classroom activity: the Human Neural Network game, where students act as neurons and pass signals forward.