Hands-on labs

Playground.

Don't just read about it — run it.

Every lab here is a real, hand-built model of the idea: the algorithm actually runs, one visible step at a time. Change the input, press play, step back, try to break it.

  • 106 interactive labs
  • 855 concepts with a lab
  • 11 topics
  • Runs in your browser
Start here

Top labs

The most hands-on labs on IKSHA: the ones with the most to drag, tweak and step through.

  1. Follow the gradient back through a tiny networkBackpropagation, Step by StepA forward pass through a 2-2-1 network, then the backward pass that applies the chain rule node by node to get for every weight and bias, and one gradient-descent update that changes the loss.You change: Inputs, target, all nine weights and biases, hidden activation, loss and learning rate.intermediate3 ways to interact4 concepts
  2. Move the kernel across the imageConvolutionintermediate3 ways to interact
  3. Train a logistic regression and move its thresholdLogistic Regressionintermediate5 ways to interact
  4. Rank pages by their linksPageRankintermediate5 ways to interact
  5. Drag points and centroids, watch k-means convergek-Means (Lloyd's Algorithm)intermediate3 ways to interact
  6. Find the cheapest routeDijkstra's Shortest Pathintermediate4 ways to interact
  7. Find the widest separating bandSupport Vector Machine: Maximum Marginadvanced4 ways to interact
  8. Partition around a pivotQuick Sortintermediate3 ways to interact
  9. Drag points, find the principal directionPrincipal Component Analysis (2-D → 1-D)intermediate4 ways to interact
  10. Send a request, read the responseHTTP Request and Responseintermediate4 ways to interact
  11. Find rows with and without an indexIndex vs Full Table Scanintermediate3 ways to interact
  12. Pick characters by positionIndexing and Slicingbeginner2 ways to interact
  13. Move down the loss curveGradient Descentbeginner2 ways to interact
106 labs

Algorithms

Watch every compare, swap and visit.

Data structures

Add, remove and follow the pointers.

How Python runs your code

Step inside real Python mechanics: loops, strings, lists, classes, files.

Databases & SQL

Run queries on a toy database and watch every row.

Networking & the web

Follow a request from DNS lookup to response.

Data formats

JSON, XML and bytes: see text become structure.

Math & bits

Flip bits and watch numbers change.

Machine learning

Drag the data, change a parameter, watch the model react.

Deep learning

Neurons, gradients and tensors, one step at a time.

Text & NLP

Turn text into tokens, ids and vectors.

Probability & randomness

Seeded experiments you can replay exactly.

Try one right now

No sign-up. Press Play, or step through with the arrow keys.

Interactive lab

Try it: Bubble Sort

How bubble sort puts a list in order by repeatedly comparing neighbouring values and swapping the pairs that are out of order.

How it works

  1. Walk the list from left to right, comparing each value with its right-hand neighbour.
  2. If the left value is larger, swap the two.
  3. After one pass the largest remaining value has 'bubbled' to the end, where it stays.
  4. Repeat on the unsorted part. A pass with no swaps means the list is sorted, so stop early.

Default run (21 steps): Unsorted array: [8, 3, 5, 1, 9]. Bubble sort compares neighbours and swaps them when they are out of order. … Sorted: [1, 3, 5, 8, 9] after 10 comparisons and 5 swaps.

Simplified: Educational version of bubble sort on up to 12 numbers. Python's built-in sort() uses a different, faster algorithm (Timsort).

Educational simulation

Loading the simulation…