Spaces:
Sleeping
Sleeping
File size: 19,095 Bytes
5e1a30c |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 |
"""
Optimization engine for the calibration system.
Implements the optimization engine component from calibration-system-spec.md
for finding optimal parameter values using various search strategies.
"""
import logging
import time
from typing import Dict, Any, List, Optional, Tuple, Callable, Iterator
from dataclasses import dataclass
from enum import Enum
import numpy as np
from itertools import product
import json
from pathlib import Path
logger = logging.getLogger(__name__)
class OptimizationStrategy(Enum):
"""Available optimization strategies."""
GRID_SEARCH = "grid_search"
BINARY_SEARCH = "binary_search"
RANDOM_SEARCH = "random_search"
GRADIENT_FREE = "gradient_free"
@dataclass
class OptimizationResult:
"""Result of parameter optimization."""
best_parameters: Dict[str, Any]
best_score: float
optimization_history: List[Dict[str, Any]]
total_evaluations: int
optimization_time: float
convergence_info: Dict[str, Any]
class OptimizationEngine:
"""
Finds optimal parameter values using various search strategies.
Implements the optimization strategies specified in calibration-system-spec.md:
- Grid search for exhaustive parameter exploration
- Binary search for single parameter optimization
- Random search for high-dimensional spaces
- Gradient-free optimization for complex parameter interactions
"""
def __init__(self, evaluation_function: Callable[[Dict[str, Any]], float]):
"""
Initialize optimization engine.
Args:
evaluation_function: Function that takes parameters and returns quality score
"""
self.evaluation_function = evaluation_function
self.optimization_history: List[Dict[str, Any]] = []
self.best_result: Optional[OptimizationResult] = None
def optimize(
self,
parameter_space: Dict[str, List[Any]],
target_metric: str = "overall_quality",
strategy: OptimizationStrategy = OptimizationStrategy.GRID_SEARCH,
max_evaluations: Optional[int] = None,
convergence_threshold: float = 0.001,
**strategy_kwargs
) -> OptimizationResult:
"""
Optimize parameters using specified strategy.
Args:
parameter_space: Dictionary mapping parameter names to lists of values
target_metric: Metric to optimize (maximize)
strategy: Optimization strategy to use
max_evaluations: Maximum number of evaluations
convergence_threshold: Convergence threshold for iterative methods
**strategy_kwargs: Additional arguments for specific strategies
Returns:
OptimizationResult with best parameters and optimization details
"""
start_time = time.time()
self.optimization_history = []
logger.info(f"Starting optimization with {strategy.value} strategy")
logger.info(f"Parameter space: {len(list(product(*parameter_space.values())))} combinations")
if strategy == OptimizationStrategy.GRID_SEARCH:
result = self._grid_search(parameter_space, target_metric, max_evaluations)
elif strategy == OptimizationStrategy.BINARY_SEARCH:
result = self._binary_search(parameter_space, target_metric, **strategy_kwargs)
elif strategy == OptimizationStrategy.RANDOM_SEARCH:
result = self._random_search(parameter_space, target_metric, max_evaluations, **strategy_kwargs)
elif strategy == OptimizationStrategy.GRADIENT_FREE:
result = self._gradient_free_search(parameter_space, target_metric, max_evaluations, convergence_threshold, **strategy_kwargs)
else:
raise ValueError(f"Unknown optimization strategy: {strategy}")
optimization_time = time.time() - start_time
# Create final result
optimization_result = OptimizationResult(
best_parameters=result["best_parameters"],
best_score=result["best_score"],
optimization_history=self.optimization_history,
total_evaluations=len(self.optimization_history),
optimization_time=optimization_time,
convergence_info=result.get("convergence_info", {})
)
self.best_result = optimization_result
logger.info(f"Optimization completed in {optimization_time:.2f}s with {len(self.optimization_history)} evaluations")
logger.info(f"Best score: {result['best_score']:.4f}")
return optimization_result
def _grid_search(
self,
parameter_space: Dict[str, List[Any]],
target_metric: str,
max_evaluations: Optional[int] = None
) -> Dict[str, Any]:
"""Exhaustive grid search over parameter space."""
param_names = list(parameter_space.keys())
param_values = list(parameter_space.values())
best_score = float('-inf')
best_params = {}
evaluations = 0
# Generate all parameter combinations
for param_combination in product(*param_values):
if max_evaluations and evaluations >= max_evaluations:
logger.info(f"Reached maximum evaluations limit: {max_evaluations}")
break
# Create parameter dictionary
current_params = dict(zip(param_names, param_combination))
# Evaluate
try:
score = self.evaluation_function(current_params)
evaluations += 1
# Track history
self.optimization_history.append({
"evaluation": evaluations,
"parameters": current_params.copy(),
"score": score,
"is_best": score > best_score
})
# Update best
if score > best_score:
best_score = score
best_params = current_params.copy()
logger.info(f"New best score: {score:.4f} with params: {current_params}")
except Exception as e:
logger.error(f"Evaluation failed for params {current_params}: {e}")
continue
return {
"best_parameters": best_params,
"best_score": best_score,
"convergence_info": {"strategy": "grid_search", "total_combinations": evaluations}
}
def _binary_search(
self,
parameter_space: Dict[str, List[Any]],
target_metric: str,
parameter_name: Optional[str] = None
) -> Dict[str, Any]:
"""Binary search for single parameter optimization."""
if len(parameter_space) != 1 and parameter_name is None:
raise ValueError("Binary search requires exactly one parameter or parameter_name to be specified")
if parameter_name:
if parameter_name not in parameter_space:
raise ValueError(f"Parameter {parameter_name} not found in parameter space")
search_param = parameter_name
other_params = {k: v[0] for k, v in parameter_space.items() if k != parameter_name}
else:
search_param = list(parameter_space.keys())[0]
other_params = {}
search_values = parameter_space[search_param]
if len(search_values) < 3:
logger.warning("Binary search requires at least 3 values, falling back to grid search")
return self._grid_search(parameter_space, target_metric)
# Sort values for binary search
sorted_values = sorted(search_values)
left, right = 0, len(sorted_values) - 1
best_score = float('-inf')
best_params = {}
evaluations = 0
while left <= right:
mid = (left + right) // 2
current_value = sorted_values[mid]
# Create parameter set
current_params = other_params.copy()
current_params[search_param] = current_value
# Evaluate
try:
score = self.evaluation_function(current_params)
evaluations += 1
self.optimization_history.append({
"evaluation": evaluations,
"parameters": current_params.copy(),
"score": score,
"search_bounds": [sorted_values[left], sorted_values[right]],
"is_best": score > best_score
})
if score > best_score:
best_score = score
best_params = current_params.copy()
logger.info(f"Binary search: new best {score:.4f} at {search_param}={current_value}")
# Decide search direction based on neighboring evaluations
if mid > 0:
left_params = other_params.copy()
left_params[search_param] = sorted_values[mid-1]
left_score = self.evaluation_function(left_params)
evaluations += 1
else:
left_score = float('-inf')
if mid < len(sorted_values) - 1:
right_params = other_params.copy()
right_params[search_param] = sorted_values[mid+1]
right_score = self.evaluation_function(right_params)
evaluations += 1
else:
right_score = float('-inf')
# Move towards better direction
if left_score > right_score:
right = mid - 1
else:
left = mid + 1
except Exception as e:
logger.error(f"Binary search evaluation failed: {e}")
break
return {
"best_parameters": best_params,
"best_score": best_score,
"convergence_info": {"strategy": "binary_search", "evaluations": evaluations}
}
def _random_search(
self,
parameter_space: Dict[str, List[Any]],
target_metric: str,
max_evaluations: int = 100,
seed: Optional[int] = None
) -> Dict[str, Any]:
"""Random search over parameter space."""
if seed is not None:
np.random.seed(seed)
param_names = list(parameter_space.keys())
param_values = list(parameter_space.values())
best_score = float('-inf')
best_params = {}
for evaluation in range(max_evaluations):
# Randomly select parameter combination
current_params = {}
for name, values in zip(param_names, param_values):
current_params[name] = np.random.choice(values)
try:
score = self.evaluation_function(current_params)
self.optimization_history.append({
"evaluation": evaluation + 1,
"parameters": current_params.copy(),
"score": score,
"is_best": score > best_score
})
if score > best_score:
best_score = score
best_params = current_params.copy()
logger.info(f"Random search: new best {score:.4f} at evaluation {evaluation + 1}")
except Exception as e:
logger.error(f"Random search evaluation failed: {e}")
continue
return {
"best_parameters": best_params,
"best_score": best_score,
"convergence_info": {"strategy": "random_search", "evaluations": max_evaluations}
}
def _gradient_free_search(
self,
parameter_space: Dict[str, List[Any]],
target_metric: str,
max_evaluations: int = 200,
convergence_threshold: float = 0.001,
population_size: int = 10
) -> Dict[str, Any]:
"""Gradient-free optimization using simple evolutionary approach."""
param_names = list(parameter_space.keys())
param_values = list(parameter_space.values())
# Initialize population
population = []
for _ in range(population_size):
individual = {}
for name, values in zip(param_names, param_values):
individual[name] = np.random.choice(values)
population.append(individual)
best_score = float('-inf')
best_params = {}
generations_without_improvement = 0
generation = 0
while len(self.optimization_history) < max_evaluations:
generation += 1
# Evaluate population
generation_scores = []
for individual in population:
try:
score = self.evaluation_function(individual)
generation_scores.append((individual, score))
self.optimization_history.append({
"evaluation": len(self.optimization_history) + 1,
"generation": generation,
"parameters": individual.copy(),
"score": score,
"is_best": score > best_score
})
if score > best_score:
improvement = score - best_score
best_score = score
best_params = individual.copy()
generations_without_improvement = 0
logger.info(f"Gradient-free: gen {generation}, new best {score:.4f}, improvement {improvement:.4f}")
if len(self.optimization_history) >= max_evaluations:
break
except Exception as e:
logger.error(f"Gradient-free evaluation failed: {e}")
continue
if not generation_scores:
break
# Selection: keep top 50%
generation_scores.sort(key=lambda x: x[1], reverse=True)
survivors = [ind for ind, _ in generation_scores[:population_size//2]]
# Reproduction: create offspring with mutations
new_population = survivors.copy()
while len(new_population) < population_size:
parent = np.random.choice(survivors)
child = parent.copy()
# Mutate random parameter
param_to_mutate = np.random.choice(param_names)
child[param_to_mutate] = np.random.choice(parameter_space[param_to_mutate])
new_population.append(child)
population = new_population
generations_without_improvement += 1
# Check convergence
if generations_without_improvement > 10: # Early stopping
logger.info(f"Gradient-free search converged after {generation} generations")
break
return {
"best_parameters": best_params,
"best_score": best_score,
"convergence_info": {
"strategy": "gradient_free",
"generations": generation,
"final_population_size": len(population),
"converged": generations_without_improvement > 10
}
}
def get_optimization_summary(self) -> str:
"""Get human-readable summary of optimization results."""
if not self.best_result:
return "No optimization completed yet."
result = self.best_result
summary = [
f"Optimization Results Summary",
f"=" * 40,
f"Best Score: {result.best_score:.4f}",
f"Total Evaluations: {result.total_evaluations}",
f"Optimization Time: {result.optimization_time:.2f}s",
f"Evaluations/Second: {result.total_evaluations/result.optimization_time:.2f}",
f"",
f"Best Parameters:"
]
for param, value in result.best_parameters.items():
summary.append(f" {param}: {value}")
if result.convergence_info:
summary.extend([
f"",
f"Convergence Info:",
f" Strategy: {result.convergence_info.get('strategy', 'unknown')}"
])
for key, value in result.convergence_info.items():
if key != 'strategy':
summary.append(f" {key}: {value}")
return "\n".join(summary)
def export_optimization_results(self, output_path: Path) -> None:
"""Export optimization results to JSON."""
if not self.best_result:
logger.error("No optimization results to export")
return
try:
export_data = {
"best_parameters": self.best_result.best_parameters,
"best_score": self.best_result.best_score,
"total_evaluations": self.best_result.total_evaluations,
"optimization_time": self.best_result.optimization_time,
"convergence_info": self.best_result.convergence_info,
"optimization_history": self.best_result.optimization_history
}
with open(output_path, 'w') as f:
json.dump(export_data, f, indent=2, default=str)
logger.info(f"Exported optimization results to {output_path}")
except Exception as e:
logger.error(f"Failed to export optimization results: {e}")
if __name__ == "__main__":
# Test optimization engine with mock evaluation function
def mock_evaluation(params):
"""Mock evaluation function for testing."""
# Simulate optimization of quadratic function
x = params.get("param_x", 0)
y = params.get("param_y", 0)
# Optimal at x=5, y=3
score = 100 - (x - 5)**2 - (y - 3)**2
return max(0, score) # Ensure non-negative
# Test parameter space
param_space = {
"param_x": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
"param_y": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
}
engine = OptimizationEngine(mock_evaluation)
# Test grid search
result = engine.optimize(
param_space,
strategy=OptimizationStrategy.GRID_SEARCH,
max_evaluations=50
)
print(engine.get_optimization_summary()) |