-
Update MYRA hybrid intelligence system
Update all model files to the latest version -
Delete correction
-
Update MYRA model and evaluation results
Finalize Hybrid Thermodynamic RBM (MYRA)
- Reconstruction: 0.98 accuracy, MSE ~0.0159
- Log-likelihood stabilized (train/test aligned)
- MCMC diagnostics: R-hat ~1.0086, tau_int ~6.9
- Energy consistency verified (gap ~17.66, entropy band valid)
- AIS estimation completed (ESS ~2483, low variance)
- Added generated and professional sample outputs -
Delete correction
-
Add SR-TRBM model implementation
- Implemented Self-Regulated Thermodynamic RBM (SR-TRBM)
- Added hybrid temperature control (micro + macro dynamics)
- Integrated Persistent Contrastive Divergence (PCD-k) training
- Introduced LLM-guided uncertainty refinement for energy corrections
- Added AIS-based partition function estimation
- Implemented multi-chain MCMC diagnostics (ESS, R-hat, autocorrelation)
- Included full evaluation pipeline (likelihood, reconstruction, entropy, diversity)
- Added visualization and diagnostic tools for energy and phase behavior
- Enabled GPU-based execution with energy tracking
This commit establishes the core experimental framework for analyzing
emergent structured behavior beyond learned statistical distributions. -
Delete correction
-
Re-upload project files (code + data)
Re-uploaded all project files, including source code and datasets.
Ensured consistency between data, preprocessing, and model components. -
Delete correction
-
Add pretrained MYRA model weights
Upload pretrained MYRA RBM weights (~190MB) for inference and evaluation.
