How sunilgentyala/qnhs-research-2026 works
QNHS is a research simulation suite evaluating a proposed quantum-neuromorphic hardware substrate, not a deployed device or application. The main workflow runs the energy, Q-STDP, MTJ-device and surface-code models and reports projected metrics; extended experiments add noisy state preparation, network learning, uncertainty and security studies. The README documents further experiment capabilities that are visible in the tree but whose implementation or wiring was not sampled, so those are represented without speculative call chains.
Main entry points: run_all_simulations.py Core runner · run_extended_experiments.py Extended runner · network_learning.py Network learning · energy_model.py Event energy
Components 16
Simulation workflows
- Core runner
run_all_simulations.py - Extended runner
run_extended_experiments.py - Extended results
results/extended_results.json
Quantum learning
- Q-STDP synapse
simulations/qstdp_simulation.py - Network learning
simulations/network_learning.py - Noisy re-preparation
simulations/reprep_noise.py - Coherent hold
simulations/coherent_hold.py - Poisoning study
simulations/poisoning.py
Device and energy
- MTJ device
simulations/mtj_resistance_model.py - Event energy
simulations/energy_model.py - Energy uncertainty
simulations/energy_uncertainty.py - Sampling pool
simulations/sampling_pool.py
Reliability and security
- Surface code
simulations/surface_code_analysis.py - MTJ random generation
simulations/mtj_trng.py - Threat model
security/threat_model.md
Other components
- Researcher
Connections 12
- Researcher → Core runner: runs
- Researcher → Extended runner: runs
- Core runner → Event energy: runs model
- Core runner → Q-STDP synapse: simulates learning
- Core runner → MTJ device: runs model
- Core runner → Surface code: runs analysis
- Extended runner → Noisy re-preparation: evaluates noise
- Extended runner → Network learning: runs networks
- Extended runner → Energy uncertainty: samples energy
- Extended runner → Extended results: writes results
- Network learning → Noisy re-preparation: uses distributions
- Event energy → MTJ device: uses device parameters