Deriva — Derive the Algorithm

← Projectsclassifier-lab / L5Trainable Artifact~28 min

one move: make an operational decision

Train one scalar weight to match a target using the same gradient-descent loop. Return the loss path instead of hiding the learning process. In the Classifier Lab system, implement this level as a deterministic contract before composing it with the next service.

Required API

def train_neuron(weight, target, learning_rate, steps):
    return [[step, weight, loss], ...]

Behavior

compute squared error against the target
move the weight down the loss gradient
record the weight and loss at every step

Examples — tap to reveal

Constraints

Use only the Python standard library.
Return deterministic values so every run can be compared with the contract.
Do not mutate caller-owned inputs.

Opening…