Interactive learning laboratory
From randomness to a network that learns.
One robot, one laboratory, six ways to decide. Inspect every calculation, train the models and see exactly what changes when a machine starts learning.
Interactive module in preparation
Same terrain, another decision
Choose the brain. Compare the outcome.
The grid, obstacles and goal stay constant. Only the way the next move is chosen changes — making the difference observable.
The Q-table changes after trials: the agent explores, then reuses decisions that worked.
Learning progression
Six levels that build intuition.
Every level answers the limits of the previous one and introduces a new family of decisions.
- 01Does not learn
Random agent
A minimal baseline that separates occasional success from progress.
Input · random direction
Mechanism · sampling - 02Does not learn
Symbolic AI
A coherent strategy written entirely by the developer.
Input · position + rules
Mechanism · explicit conditions - 03Learns
Q-learning
The robot improves decisions through trials, rewards and penalties.
Input · state + reward
Mechanism · Q-table update - 04Learns
Perceptron
A first decision boundary built from numerical inputs.
Input · inputs + weights
Mechanism · supervised - 05Fixed weights
Multilayer network
A hidden layer demonstrates forward propagation and intermediate representations.
Input · activations
Mechanism · forward propagation - 06Learns
Backpropagation
Loss produces gradients that progressively correct the weights.
Input · error + gradients
Mechanism · gradient descent
Learn by manipulating
A decision becomes understandable when you can follow it.
SupraLab AI connects the robot’s visible movement to the values, weights, probabilities and errors that produced it.
- 01
Simulate
Move step by step or let the environment run at several speeds.
- 02
Train
Accelerate Q-learning and train the supervised models.
- 03
Inspect
Read inputs, activations, Q-values, gradients and losses.
- 04
Compare
Separate an occasional success from a strategy that improves.
Actual scope
A readable experiment for learning.
The lab prioritises intuition and inspection. Models are small, the map is fixed and supervised data comes from the experiment’s environment.
- Six approaches compared under the same conditions
- Calculations and decisions visible as they act
- No claim of industrial performance
Inside the ecosystem
After the mechanism, return to mastery.
An intuition observed in the Lab can become a concept to understand, practise and retrieve in FlashLearning.
Product access
Follow the SupraLab AI launch.
Receive the interactive experience link when it goes live.