AutoML Pipeline Leaderboard: Successive Halving Race (2D)

Automated Machine Learning has to solve two problems at once: which algorithm family to use, and which hyperparameters to give it — the Combined Algorithm Selection and Hyperparameter optimization (CASH) problem. Training every candidate pipeline to full convergence is far too expensive, so real AutoML systems race a whole portfolio of pipelines under a Successive-Halving budget schedule: everyone gets a small, noisy evaluation budget, the weakest performers are eliminated each round, and the survivors get a bigger budget on the next round. This 2D simulator renders that race as a bar-chart leaderboard — one bar per candidate pipeline, colored by algorithm family, rising and falling with its live score estimate — with real controls over the portfolio size, the halving aggressiveness and the evaluation noise, plus drag-to-pan and scroll-to-zoom so you can inspect the field closely.