Spaces:
Runtime error
Runtime error
File size: 10,900 Bytes
12af533 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 |
#!/usr/bin/env python3
"""
Layer 2: Self-Evolution Engine
Auto-Skill-Detector, Pattern-Detection-Engine, Q-Score-Optimizer
From skill-ecosystem-integration-guide.md
"""
import sys
sys.path.append('/home/claude')
from layer0_foundation import Skill, build_adjacency_matrix, synthesize_skills
from layer1_skills import create_layer1_ecosystem
import numpy as np
from typing import List, Dict, Tuple
from collections import defaultdict
from dataclasses import dataclass
import json
# ============================================================================
# AUTO-SKILL-DETECTOR (Q=0.952)
# ============================================================================
class AutoSkillDetector:
"""
Monitors patterns and generates skills automatically.
Trigger: 3+ repetitions OR complexity > 0.8
"""
def __init__(self, threshold: int = 3):
self.pattern_cache = defaultdict(int)
self.threshold = threshold
self.generated_skills = []
self.skill = Skill(
name="Auto-Skill-Detector",
G=0.95, C=0.93, S=0.96, A=0.95, H=0.96, V=0.95, P=0.95, T=0.92
)
def record_usage(self, skills_used: List[str], task: str, outcome: float):
"""Track skill co-usage patterns"""
pattern_sig = "+".join(sorted(skills_used))
self.pattern_cache[pattern_sig] += 1
# Check if threshold reached
if self.pattern_cache[pattern_sig] >= self.threshold:
return self._should_generate(pattern_sig, skills_used)
return None
def _should_generate(self, pattern_sig: str, skills_used: List[str]) -> bool:
"""Check if we should generate new skill"""
# Don't regenerate
if pattern_sig in [s.name for s in self.generated_skills]:
return False
# Calculate modularity (simplified)
modularity = len(skills_used) / 10.0 # More skills = higher modularity
if modularity > 0.2:
return True
return False
def generate_skill(self, parent_skills: List[Skill]) -> Skill:
"""Generate emergent skill from pattern"""
emergent = synthesize_skills(parent_skills, gamma=0.20)
# Validate Q-score
if emergent.q_score() >= 0.75:
self.generated_skills.append(emergent)
return emergent
return None
# ============================================================================
# PATTERN-DETECTION-ENGINE (Q=0.941)
# ============================================================================
class PatternDetectionEngine:
"""
Spectral graph analysis for pattern detection.
Complexity: O(n log n)
"""
def __init__(self):
self.skill = Skill(
name="Pattern-Detection-Engine",
G=0.94, C=0.92, S=0.95, A=0.96, H=0.94, V=0.96, P=0.95, T=0.93
)
def detect_communities(self, skills: List[Skill]) -> Dict:
"""
Spectral clustering via Laplacian eigendecomposition.
L = D - A (Laplacian)
Eigendecomposition: L v = λ v
"""
n = len(skills)
if n < 2:
return {"communities": [], "modularity": 0.0}
# Build adjacency matrix
A = build_adjacency_matrix(skills)
# Degree matrix
D = np.diag(A.sum(axis=1))
# Laplacian
L = D - A
# Eigendecomposition
eigenvalues, eigenvectors = np.linalg.eigh(L)
# Use k smallest non-zero eigenvalues for clustering
k = min(3, n - 1) # Number of communities
# Simple clustering based on eigenvector signs
communities = []
for i in range(k):
if i < len(eigenvalues) - 1:
community = np.where(eigenvectors[:, i+1] > 0)[0].tolist()
if len(community) > 0:
communities.append(community)
# Calculate modularity
modularity = self._compute_modularity(A, communities)
return {
"communities": communities,
"modularity": modularity,
"significant": modularity > 0.3
}
def _compute_modularity(self, A: np.ndarray, communities: List[List[int]]) -> float:
"""
Q_modularity = (1/2m) Σᵢⱼ [Aᵢⱼ - (kᵢkⱼ/2m)] δ(cᵢ, cⱼ)
"""
m = A.sum() / 2 # Total edges
if m == 0:
return 0.0
n = A.shape[0]
k = A.sum(axis=1) # Degrees
Q = 0.0
for community in communities:
for i in community:
for j in community:
Q += A[i, j] - (k[i] * k[j]) / (2 * m)
Q /= (2 * m)
return Q
# ============================================================================
# Q-SCORE-OPTIMIZER (Q=0.928)
# ============================================================================
class QScoreOptimizer:
"""
Continuous quality improvement via gradient-based optimization.
Guarantee: Q(skill_t+1) ≥ Q(skill_t)
"""
def __init__(self, target_q: float = 0.90, epsilon: float = 0.001):
self.target_q = target_q
self.epsilon = epsilon
self.skill = Skill(
name="Q-Score-Optimizer",
G=0.93, C=0.91, S=0.94, A=0.94, H=0.93, V=0.92, P=0.93, T=0.92
)
# Dimension weights for optimization priority
self.weights = {
'G': 0.18, 'C': 0.20, 'S': 0.18, 'A': 0.16,
'H': 0.12, 'V': 0.08, 'P': 0.05, 'T': 0.03
}
def optimize(self, skill: Skill, max_iterations: int = 5) -> Tuple[Skill, Dict]:
"""Iterative improvement until target Q reached"""
history = []
current = skill
for iteration in range(max_iterations):
q_current = current.q_score()
history.append(q_current)
# Check convergence
if q_current >= self.target_q:
return current, {
"converged": True,
"iterations": iteration + 1,
"final_q": q_current,
"history": history
}
# Find bottleneck
bottleneck = self._find_bottleneck(current)
# Improve bottleneck
current = self._improve_dimension(current, bottleneck)
# Check monotonic improvement
q_new = current.q_score()
assert q_new >= q_current, "Non-monotonic improvement!"
# Check if change is too small
if abs(q_new - q_current) < self.epsilon:
return current, {
"converged": True,
"iterations": iteration + 1,
"final_q": q_new,
"history": history,
"reason": "epsilon_convergence"
}
return current, {
"converged": False,
"iterations": max_iterations,
"final_q": current.q_score(),
"history": history
}
def _find_bottleneck(self, skill: Skill) -> str:
"""Find dimension with lowest weighted score"""
dimensions = {
'G': skill.G, 'C': skill.C, 'S': skill.S, 'A': skill.A,
'H': skill.H, 'V': skill.V, 'P': skill.P, 'T': skill.T
}
# Calculate weighted scores
weighted_scores = {
dim: score * self.weights[dim]
for dim, score in dimensions.items()
}
# Find minimum
bottleneck = min(weighted_scores, key=weighted_scores.get)
return bottleneck
def _improve_dimension(self, skill: Skill, dimension: str, delta: float = 0.05) -> Skill:
"""Improve specific dimension"""
improved = Skill(
name=skill.name,
G=skill.G, C=skill.C, S=skill.S, A=skill.A,
H=skill.H, V=skill.V, P=skill.P, T=skill.T,
priority=skill.priority, cost=skill.cost
)
# Improve dimension (clip to [0, 1])
current_value = getattr(improved, dimension)
new_value = min(1.0, current_value + delta)
setattr(improved, dimension, new_value)
return improved
# ============================================================================
# TESTING
# ============================================================================
if __name__ == "__main__":
print("=" * 60)
print("LAYER 2: SELF-EVOLUTION ENGINE TESTS")
print("=" * 60)
# Get Layer 1 skills
ecosystem = create_layer1_ecosystem()
layer1_skills = ecosystem["skills"]
print("\n[Test 1] Auto-Skill-Detector")
detector = AutoSkillDetector(threshold=3)
# Simulate usage patterns
for i in range(5):
result = detector.record_usage(
["Meta-Learning", "Transfer-Learning"],
f"task_{i}",
0.9
)
if i >= 2:
print(f" Iteration {i+1}: Should generate = {result}")
# Generate skill
if detector._should_generate("Meta-Learning+Transfer-Learning", ["Meta-Learning", "Transfer-Learning"]):
new_skill = detector.generate_skill([layer1_skills[0], layer1_skills[1]])
if new_skill:
print(f" Generated: {new_skill.name}, Q={new_skill.q_score():.3f}")
print("\n[Test 2] Pattern-Detection-Engine")
pattern_engine = PatternDetectionEngine()
result = pattern_engine.detect_communities(layer1_skills)
print(f" Communities found: {len(result['communities'])}")
print(f" Modularity: {result['modularity']:.3f}")
print(f" Significant: {result['significant']}")
print("\n[Test 3] Q-Score-Optimizer")
optimizer = QScoreOptimizer(target_q=0.95)
# Create suboptimal skill
test_skill = Skill(
name="Test-Skill",
G=0.7, C=0.75, S=0.8, A=0.7, H=0.75, V=0.7, P=0.6, T=0.65
)
print(f" Initial Q: {test_skill.q_score():.3f}")
optimized, info = optimizer.optimize(test_skill)
print(f" Final Q: {optimized.q_score():.3f}")
print(f" Iterations: {info['iterations']}")
print(f" Converged: {info['converged']}")
print(f" Improvement: +{optimized.q_score() - test_skill.q_score():.3f}")
# Verify monotonic improvement
print(f" History: {[f'{q:.3f}' for q in info['history']]}")
for i in range(len(info['history']) - 1):
assert info['history'][i+1] >= info['history'][i], "Non-monotonic!"
print(f" ✅ Monotonic improvement verified")
print("\n" + "=" * 60)
print("LAYER 2 TESTS COMPLETE")
print("=" * 60)
|