Science Advances

This PDF file includes:

• Supplementary Text
  
• section S1. Analysis of the estimators RPLE, RISE, and logRISE
  
• section S2. On optimization techniques for minimizing the estimators
  
• section S3. On the procedure for M* selection
  
• section S4. On the theoretical predictions for λ selection
  
• section S5. Empirical selection of the regularization parameter λ
  
• section S6. Hyperparameter λ selection through cross-validation
  
• section S7. Scalings of the RISE with respect to γ
  
• section S8. High-temperature expansion of the RISE and the RPLE
  
• fig. S1. Schematic representation of the proof strategy for bounding the reconstruction error.
  
• fig. S2. Dependence of the number of samples M* required for a correct structure recovery on the regularization coefficient.
  
• fig. S3. Selection of the hyperparameter λ through the K-fold cross-validation method on a 4 × 4 spin glass system on a square lattice.
  
• fig. S4. Values of M* and γ exponents for the RISE across different test cases.
  
• References (42–45)

Download PDF