How a neuron learns to lend

An old prediction problem, a new way of solving it.

Income Debt load Σ + b σ(z) Line thickness = weight size. Gray line = negative weight.
● repaid   ■ defaulted   Shading = neuron's current verdict (green approve, orange decline). Click an applicant to send them through the neuron.
Weight w₁ Weight w₂ Bias b
Applicant Predicted chance of repaying Ledger cases called correctly Loss (how wrong, on average)