SPARQ — Benchmark Data and Trained Models

# SPARQ — Benchmark Data and Trained Models Companion dataset for the manuscript: > **SPARQ — spiking physics-in-the-loop autonomous reinforcement triage> of quantum emitters: a closed-loop, event-driven machine-learning> framework for characterizing solid-state single-photon sources**> T. M. Mahim, M. N. Islam, M. M. Rahman, A. S. M. Mohsin This deposit contains the complete benchmark measurements underlyingevery figure and number in the article, and all trained modelcheckpoints. It does **not** contain source code. ## Contents ```benchmark_data/ exp1_validation.json twin validation vs. exact master-equation g2 (MC curves, chi^2, twin-vs-MC means, Fano) exp2_estimators.json estimator accuracy/MAE sweeps vs. acquisition time (fit, CNNs, SNN; 5 seeds), MC-Bayes reference, time-to-target exp2b_snn.json activity-regularized SNN sweeps, anytime latency table, measured energy accounting exp3_adjoint.json adjoint protocol trajectory, per-budget accuracies, profile Fisher-information sweep exp4_gan.json real-data (sps-quality) series references, WGAN-GP training log, early-estimation MAEs exp4c_floor.json in-domain information-floor MAE exp5_rl.json triage baselines, SAC learning curves and final evaluations (PER vs. uniform) exp5_dwell.json learned policy dwell allocations exp5b_oracle.json clairvoyant oracle-stopping bound exp6_graph.json naive two-platform transfer ablation exp6b_graph.json platform-randomized graph transfer resultstrained_models/ PyTorch state_dicts cnn_pitl.pt physics-in-the-loop CNN estimator cnn_clean.pt clean-trained ablation CNN cnn_triage.pt triage estimator (purity + good + regression) cnn_graph_syn.pt graph-conditioned estimator (synthetic-trained) snn_pitl.pt spiking estimator (PITL) snn_sparse.pt activity-regularized spiking estimator gan_G.pt WGAN-GP generator (twin refiner) sac_pi.pt SAC+PER triage policy network``` All JSON files are self-describing (keys named after the quantities inthe article). Model architectures are specified in the manuscript'ssupplementary material. The experimental quantum-dot HBT measurements referenced by`exp4_gan.json` are NOT redistributed here; they are openly availablefrom the UTS-CASLab *sps-quality* repository(https://github.com/UTS-CASLab/sps-quality, Kedziora et al., Mach.Learn.: Sci. Technol. 4, 045042 (2023)). ## License CC-BY 4.0.