| """Plot recorded V2 results without changing any experiments.""" |
| from pathlib import Path |
| import json |
| import matplotlib |
| matplotlib.use('Agg') |
| import matplotlib.pyplot as plt |
|
|
| root=Path(__file__).resolve().parents[1] |
| s=json.loads((root/'results/v2_program_summary.json').read_text()) |
| out=root/'figures';out.mkdir(exist_ok=True) |
| plt.rcParams.update({'font.family':'DejaVu Sans','font.size':9,'pdf.fonttype':42, |
| 'axes.spines.top':False,'axes.spines.right':False}) |
| names={'corrected':'Learned proposer + exact compiler','frozen_proposer':'Frozen proposer', |
| 'no_feedback':'No-feedback lineage','direct_tournament':'Direct two-order tournament'} |
| colors={'corrected':'#167a64','frozen_proposer':'#7a698e','no_feedback':'#9b6a51','direct_tournament':'#648498'} |
| fig,axes=plt.subplots(1,2,figsize=(11,4.3)) |
| for method,label in names.items(): |
| r=[x for x in s['aggregate'] if x['method']==method] |
| g=[x['generation'] for x in r] |
| axes[0].plot(g,[x['mean_proposed_heldout_loss_ratio'] for x in r],'-o',color=colors[method],label=label,markersize=4) |
| axes[1].plot(g,[x['mean_heldout_loss_ratio'] for x in r],'-o',color=colors[method],label=label,markersize=4) |
| for ax in axes: |
| ax.set_xlabel('Proposal generation');ax.grid(alpha=.18);ax.set_xticks(range(1,6)) |
| ax.set_ylabel('Held-out residual loss ratio; lower is better') |
| axes[0].set_title('Mean proposal quality improves') |
| axes[1].set_title('Best found: frozen proposer remains competitive') |
| axes[0].legend(fontsize=7.8,loc='upper right') |
| fig.suptitle('Bounded evidence-driven program proposal',fontsize=14) |
| fig.text(.5,-.01,'8 systems, 24 held-out initial states each; 240 training program executions per method and system. Fixed search algorithm.',ha='center',fontsize=8) |
| fig.tight_layout() |
| for ext in ['png','pdf']: |
| fig.savefig(out/f'v2_proposal_results.{ext}',dpi=180,bbox_inches='tight') |
| plt.close(fig) |
| print('Wrote v2_proposal_results.pdf and .png') |
|
|
