Taught in Base Camp · Unit 8
How machines learn from data
Features, labels, training and testing: how a model finds a pattern in examples instead of following rules someone wrote down.
For parents · Classes 7 to 10
Why does a chatbot give a different answer each time? How does an app learn to tell things apart? Your child finds out by playing with working models, writing a little Python and testing what they know against the national AI Olympiad’s syllabus.
Try it now · k-nearest neighbours
One simple way: look at the closest examples it already knows, and let them vote. Tap or drag the ringed point on the plot, then change k, the number of neighbours that get a vote.
What they’ll understand
Each one has a short guide to read and a simulator to play with, all free. Open any of them now.
Taught in Base Camp · Unit 8
Features, labels, training and testing: how a model finds a pattern in examples instead of following rules someone wrote down.
Taught in Base Camp · Unit 9
Why high accuracy can hide a useless model, why a model can memorise instead of learn, and how to test one fairly.
Taught in Base Camp · Unit 7Summit · Unit 11
Tokens, chances and temperature: why replies vary, why fluent text is not the same as a checked fact, and how to write a clear prompt.
Taught in Base Camp · Unit 7Summit · Unit 11
How one model can treat two groups differently, why deleting a column doesn’t delete the bias, and what a fairer threshold costs.
Guide Simulator
How they learn
Short sessions on most days. Even the short video lessons stop every minute or two to ask a question: your child does something on every screen.
Move a slider, drag a point, toss a coin a thousand times. Each one shows an idea working, and each is free.
The idea in plain words, a worked example and the usual traps, with a few check questions at the end.
Real Python from Base Camp Unit 5, with nothing to install. Code is checked automatically against hidden tests, so your child knows at once whether it works.
Original questions, every one with a worked explanation, at three tiers: Trail, Ridge and Peak.
The path builds up to questions on the national AI Olympiad’s syllabus, so understanding has a clear, demanding measure.
The yardstick
“Learning AI” can mean anything. A national syllabus, written for school students, makes it specific.
Set by ACM India
It is the syllabus of the Indian National AI Olympiad (INAIO), run by ACM India as India’s route to the International Olympiad in Artificial Intelligence (IOAI).
It tests understanding, not tool use
The syllabus covers logic, probability, algorithms and how models are trained and measured, then hands-on Python with data. Stage 1 allows no AI tools.
No pressure to compete
INAIO is open to Classes 9 to 12, and taking part is your family’s choice. The syllabus is a useful map either way.
A question on the simulator you just tried. Nearest neighbours is one of the methods on the INAIO Stage 1 syllabus. Pick an answer to see the explanation.
Try one · Base Camp · Unit 8 · How machines learn
Why can k = 3 nearest neighbours be safer than k = 1?
The offer
Base Camp is the first year: 12 units from puzzles and Python to how machines learn. One payment plus GST, with no subscription.
Free
With a free account, unless it says otherwise.
Year 1 · Foundations
Classes 7–10 starting from scratch
₹14,999+ GST · one-time · 12 months
₹17,699 with 18% GST · about ₹1,250 a month + GST
Paid plans can be bought here shortly. Start free now: everything you do carries over.
Starting in Class 7 or 8? Head Start covers Base Camp and then Summit for ₹29,999 + GST, with 24 months of access.
Free to start
Start with Unit 1 of Base Camp on a free account. Every simulator and guide is free too, and there is nothing to install.
No payment needed · No adverts
k-NN does no training at all. To classify a new point it finds the k closest labelled points and lets them vote. Small k follows every wobble in the data (jagged borders, sensitive to noise). Large k smooths the borders but can drown out small groups.
What counts as “close” depends on the distance metric. Euclidean measures in a straight line, so the neighbourhood is a circle. Manhattan adds up horizontal and vertical steps, so it is a diamond. With two classes and an even k, the vote can tie, which is why people usually pick an odd k.
The 5 nearest points lie within 1.13 units (euclidean). Majority wins.
k-NN does no training at all. To classify a new point it finds the k closest labelled points and lets them vote. Small k follows every wobble in the data (jagged borders, sensitive to noise). Large k smooths the borders but can drown out small groups.
What counts as “close” depends on the distance metric. Euclidean measures in a straight line, so the neighbourhood is a circle. Manhattan adds up horizontal and vertical steps, so it is a diamond. With two classes and an even k, the vote can tie, which is why people usually pick an odd k.
Things to try