HomeAI & Machine LearningGradient Boosting 2D: Residual Heatmap Across Rounds

Gradient Boosting 2D: Residual Heatmap Across Rounds

2D companion to the gradient-boosting simulator: a scrolling round-by-point residual heatmap and a live tree-partition strip, driven by the same real greedy SSE-minimizing regression trees, shrinkage and residual fitting — computed independently as a grid, not a flattened 3D view.

AI & Machine Learning2DModerate60 FPS📱 Mobile-adapted⇄ 3D version
2d-gradient-boosting-decision-tree-ensemble ↗ Open standalone

This is the 2D companion to the 3D gradient-boosting simulator, computed independently rather than rendered as a flattened camera angle on the same scene. It runs the identical algorithm — a real greedy, SSE-minimizing regression tree fit to residuals every round, added back with shrinkage — but shows it through two grid-native panels: a scrolling round-by-point residual heatmap that cools from red/blue toward white as the ensemble learns, and a live strip of the most recent tree's x-partition with its per-leaf residual bars. Shrinkage, tree depth and dataset noise remain live controls, so the same bias/variance trade-off is visible directly in how fast — and how evenly — the heatmap converges.

⚙ Under the hood

2D companion to the gradient-boosting simulator: a scrolling round-by-point residual heatmap and a live tree-partition strip, driven by the same real greedy SSE-minimizing regression trees, shrinkage and residual fitting as the 3D version — rendered as a grid, not a flattened camera view.

gradient boostingdecision treesensemble learningregressionmachine learningresidual fittingheatmap

2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install

What did you find?

Add reproduction steps (optional)