Datasets:
Formats:
json
Languages:
English
Size:
< 1K
Tags:
time-series
time-series-decomposition
benchmark
component-recovery
symbolic-regression
icml-2026
License:
File size: 3,226 Bytes
17b7ba4 | 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 | import numpy as np
from typing import Dict, Any, Tuple, Optional
from ..core import DecompResult
from ..registry import MethodRegistry
from .utils import dominant_frequency
try:
from PyEMD import CEEMDAN
_HAS_CEEMDAN = True
except ImportError:
_HAS_CEEMDAN = False
def estimate_imf_dom_freqs(imfs: np.ndarray, fs: float) -> np.ndarray:
freqs = []
for imf in imfs:
freqs.append(dominant_frequency(imf, fs))
return np.array(freqs, dtype=float)
def assign_imfs_to_components(
freqs: np.ndarray,
primary_freq: Optional[float],
config: Dict[str, Any],
) -> Tuple[np.ndarray, np.ndarray]:
trend_threshold = float(config.get("trend_freq_max", 0.02))
if primary_freq:
trend_threshold = config.get("trend_freq_max", primary_freq / 8.0)
season_band_factor = float(config.get("season_band_factor", 2.0))
trend_mask = freqs <= max(trend_threshold, 1e-6)
season_mask = np.zeros_like(freqs, dtype=bool)
if primary_freq and primary_freq > 0:
low = primary_freq / max(season_band_factor, 1.0)
high = primary_freq * max(season_band_factor, 1.0)
season_mask = (freqs >= low) & (freqs <= high)
if not trend_mask.any():
trend_mask[np.argmin(freqs)] = True
if not season_mask.any():
if primary_freq:
idx = int(np.argmin(np.abs(freqs - primary_freq)))
season_mask[idx] = True
else:
season_mask[~trend_mask] = True
return trend_mask, season_mask
@MethodRegistry.register("CEEMDAN")
def ceemdan_decompose(
y: np.ndarray,
params: Dict[str, Any],
) -> DecompResult:
if not _HAS_CEEMDAN:
raise ImportError("PyEMD>=1.0 is required for CEEMDAN decomposition.")
cfg = params.copy()
ceemdan = CEEMDAN()
if "trials" in cfg:
ceemdan.trials = int(cfg["trials"])
if "noise_width" in cfg:
ceemdan.noise_width = float(cfg["noise_width"])
imfs = ceemdan(y)
imfs = np.asarray(imfs, dtype=float)
if imfs.ndim == 1:
imfs = imfs[np.newaxis, :]
if imfs.size == 0:
zeros = np.zeros_like(y)
return DecompResult(
trend=zeros,
season=zeros,
residual=y.copy(),
meta={"method": "CEEMDAN", "imfs": []}
)
fs = float(cfg.get("fs", 1.0))
freqs = estimate_imf_dom_freqs(imfs, fs)
primary_period = cfg.get("primary_period")
primary_freq = 1.0 / float(primary_period) if primary_period else None
trend_mask, season_mask = assign_imfs_to_components(freqs, primary_freq, cfg)
trend = imfs[trend_mask].sum(axis=0) if trend_mask.any() else np.zeros_like(y)
season = imfs[season_mask].sum(axis=0) if season_mask.any() else np.zeros_like(y)
residual = y - trend - season
return DecompResult(
trend=trend,
season=season,
residual=residual,
meta={
"method": "CEEMDAN",
"dominant_frequencies": freqs.tolist(),
"trend_mask": trend_mask.tolist(),
"season_mask": season_mask.tolist(),
"primary_period": primary_period,
}
)
|