Why does an imitation learner struggle after its first mistake?
Deterministic tabular dataset aggregation in an eight-step offset simulator; no neural training, driving deployment, or claim to reproduce the paper benchmarks.
This notebook runs the complete saved experiment, checks its assertions, and regenerates the SVG from the computed results. Python 3.10+ and standard-library modules only. Place experiment.py and figure_generation.py beside the notebook. The exact loaded assets are recorded in notebook metadata.
iteration=0 error=3 trace=[0, 0, 0, 1, 1, 1, 0, 0, 0]
visited=[0, 1]
iteration=1 error=4 trace=[0, 0, 0, 1, 0, 0, -1, -1, -1]
visited=[-1, 0, 1]
iteration=2 error=2 trace=[0, 0, 0, 1, 0, 0, -1, 0, 0]
visited=[-1, 0, 1]
Aggregated table: {-1: 1, 0: 0, 1: -1}
Centered test error=2
Unseen offset-2 test error=19
table round=0, exhaustive centered mean error=6.9712, unknown-state decisions=60.4938%
table round=1, exhaustive centered mean error=5.9300, unknown-state decisions=41.1523%
table round=2, exhaustive centered mean error=4.0000, unknown-state decisions=0.0000%
Regenerated figure.svg from results.json.
Inspect the saved result
The article's static figure is available before running code. The regenerated figure.svg uses results.json, and every plotted quantity comes from the experiment.
Change an input in experiment.py, rerun both code cells, then compare the saved result and figure.
Python runs locally in this page. Saved results and figures are visible before execution.