Notification Channel Optimizer 2D — Multi-Armed Bandit Bar Race
A flat 2D multi-armed bandit dashboard: watch epsilon-greedy and UCB1 learn live, message by message, which channel (SMS, email, app push, mail) most reliably reaches citizens, with a growing bar race, a confidence-bound overlay and a cumulative-regret time chart.
A government agency needs to send a citizen alert, but which channel — SMS, email, app push, or mail — will actually get read? This 2D dashboard runs a real multi-armed bandit AI that sends one synthetic message at a time, observes whether it was "opened" against a hidden per-channel true rate, and updates its running belief. Switch between epsilon-greedy and UCB1 selection strategies and watch the four channel bars race and settle, a translucent confidence-bound cap track each estimate under UCB1, and a live cumulative-regret chart trace how efficiently the AI is converging on the best-performing channel.
A flat 2D multi-armed bandit dashboard: watch epsilon-greedy and UCB1 learn live, message by message, which channel (SMS, email, app push, mail) most reliably reaches citizens, with a growing bar race, a confidence-bound overlay and a cumulative-regret time chart.
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install