Cross-Domain Literature Survey

K-Means Clustering of All Constructs

engine kmeans_clustering

config (3 keys)

{
  "n_clusters": 7,
  "n_clusters_note": "Start with 7 (one per domain) and examine whether natural clusters align with domain boundaries",
  "variables": [
    "formation_energy_per_atom",
    "energy_above_hull",
    "revised_tolerance_factor",
    "perovskite_formability",
    "intercalation_voltage",
    "cathode_capacity",
    "thermodynamic_selectivity",
    "synthesis_success_rate",
    "oxygen_evolution_overpotential",
    "persistent_homology_descriptor",
    "discovery_acceleration_factor",
    "neural_network_potential"
  ]
}

expected results (1 key)

{
  "cluster_alignment_with_domains": {
    "note": "If clusters match domains, constructs are domain-specific; if they cut across, shared structure exists"
  }
}

Cross-Domain Correlation Matrix

engine correlation_matrix

config (1 key)

{
  "variables": [
    "formation_energy_per_atom",
    "energy_above_hull",
    "revised_tolerance_factor",
    "intercalation_voltage",
    "thermodynamic_selectivity",
    "oxygen_evolution_overpotential",
    "persistent_homology_descriptor",
    "discovery_acceleration_factor",
    "neural_network_potential",
    "energy_prediction_mae",
    "synthesis_success_rate"
  ]
}

expected results (3 keys)

{
  "discovery_acceleration_factor↔neural_network_potential": {
    "direction": "positive",
    "note": "Better ML models accelerate discovery"
  },
  "formation_energy_per_atom↔intercalation_voltage": {
    "direction": "negative",
    "note": "More negative formation energy → higher voltage (thermodynamic relationship)"
  },
  "neural_network_potential↔energy_prediction_mae": {
    "direction": "negative",
    "note": "Better NNPs yield lower MAE"
  }
}

OLS: What Drives Discovery Acceleration?

engine ols_regression←cross_domain_correlations

config (3 keys)

{
  "diagnostics": true,
  "predictors": [
    "neural_network_potential",
    "thermodynamic_selectivity",
    "energy_prediction_mae",
    "synthesis_success_rate"
  ],
  "target": "discovery_acceleration_factor"
}

expected results (3 keys)

{
  "neural_network_potential_coefficient": {
    "direction": "positive"
  },
  "r_squared": {
    "min": 0.4,
    "note": "Exploratory model — moderate R² expected given construct heterogeneity"
  },
  "synthesis_success_rate_coefficient": {
    "direction": "positive"
  }
}