鱼皮AI导航
鱼皮AI导航
  • 主页
  • 社区
  • AI工具
  • AI知识库
  • AI提示词
  • MCP
  • Skills
  • AI学习资源
  • AI资讯
  • GitHub开源
登录/注册

box-least-squares

专门用于探测光变曲线中周期性箱型凹陷信号,适用于系外行星凌星和食双星事件的识别,通过拟合周期、持续时间、深度和参考时刻等参数实现高灵敏度周期搜索。

研究
访问 GitHub
文档Skills.MD
name
box-least-squares
description
Box Least Squares (BLS) periodogram for detecting transiting exoplanets and eclipsing binaries. Use when searching for periodic box-shaped dips in light curves. Alternative to Transit Least Squares, available in astropy.timeseries. Based on Kovács et al. (2002).

Box Least Squares (BLS) Periodogram

The Box Least Squares (BLS) periodogram is a statistical tool for detecting transiting exoplanets and eclipsing binaries in photometric time series data. BLS models a transit as a periodic upside-down top hat (box shape) and finds the period, duration, depth, and reference time that best fit the data.

Overview

BLS is built into Astropy and provides an alternative to Transit Least Squares (TLS). Both search for transits, but with different implementations and performance characteristics.

Key parameters BLS searches for:

  • Period (orbital period)
  • Duration (transit duration)
  • Depth (how much flux drops during transit)
  • Reference time (mid-transit time of first transit)

Installation

BLS is part of Astropy:

▼
bash
复制代码
pip install astropy

Basic Usage

▼
python
复制代码
import numpy as np import astropy.units as u from astropy.timeseries import BoxLeastSquares # Prepare data # time, flux, and flux_err should be numpy arrays or Quantities t = time * u.day # Add units if not already present y = flux dy = flux_err # Optional but recommended # Create BLS object model = BoxLeastSquares(t, y, dy=dy) # Automatic period search with specified duration duration = 0.2 * u.day # Expected transit duration periodogram = model.autopower(duration) # Extract results best_period = periodogram.period[np.argmax(periodogram.power)] print(f"Best period: {best_period:.5f}")

Using autopower vs power

autopower: Automatic Period Grid

Recommended for initial searches. Automatically determines appropriate period grid:

▼
python
复制代码
# Specify duration (or multiple durations) duration = 0.2 * u.day periodogram = model.autopower(duration) # Or search multiple durations durations = [0.1, 0.15, 0.2, 0.25] * u.day periodogram = model.autopower(durations)

power: Custom Period Grid

For more control over the search:

▼
python
复制代码
# Define custom period grid periods = np.linspace(2.0, 10.0, 1000) * u.day duration = 0.2 * u.day periodogram = model.power(periods, duration)

Warning: Period grid quality matters! Too coarse and you'll miss the true period.

Objective Functions

BLS supports two objective functions:

1. Log Likelihood (default)

Maximizes the statistical likelihood of the model fit:

▼
python
复制代码
periodogram = model.autopower(0.2 * u.day, objective='likelihood')

2. Signal-to-Noise Ratio (SNR)

Uses the SNR with which the transit depth is measured:

▼
python
复制代码
periodogram = model.autopower(0.2 * u.day, objective='snr')

The SNR objective can improve reliability in the presence of correlated noise.

Complete Example

▼
python
复制代码
import numpy as np import matplotlib.pyplot as plt import astropy.units as u from astropy.timeseries import BoxLeastSquares # Load and prepare data data = np.loadtxt('light_curve.txt') time = data[:, 0] * u.day flux = data[:, 1] flux_err = data[:, 3] # Create BLS model model = BoxLeastSquares(time, flux, dy=flux_err) # Run BLS with automatic period grid # Try multiple durations to find best fit durations = np.linspace(0.05, 0.3, 10) * u.day periodogram = model.autopower(durations, objective='likelihood') # Find peak max_power_idx = np.argmax(periodogram.power) best_period = periodogram.period[max_power_idx] best_duration = periodogram.duration[max_power_idx] best_t0 = periodogram.transit_time[max_power_idx] max_power = periodogram.power[max_power_idx] print(f"Period: {best_period:.5f}") print(f"Duration: {best_duration:.5f}") print(f"T0: {best_t0:.5f}") print(f"Power: {max_power:.2f}") # Plot periodogram import matplotlib.pyplot as plt plt.plot(periodogram.period, periodogram.power) plt.xlabel('Period [days]') plt.ylabel('BLS Power') plt.show()

Peak Statistics for Validation

Use compute_stats() to calculate detailed statistics about a candidate transit:

▼
python
复制代码
# Get statistics for the best period stats = model.compute_stats( periodogram.period[max_power_idx], periodogram.duration[max_power_idx], periodogram.transit_time[max_power_idx] ) # Key statistics for validation print(f"Depth: {stats['depth']:.6f}") print(f"Depth uncertainty: {stats['depth_err']:.6f}") print(f"SNR: {stats['depth_snr']:.2f}") print(f"Odd/Even mismatch: {stats['depth_odd'] - stats['depth_even']:.6f}") print(f"Number of transits: {stats['transit_count']}") # Check for false positives if abs(stats['depth_odd'] - stats['depth_even']) > 3 * stats['depth_err']: print("Warning: Significant odd-even mismatch - may not be planetary")

Validation criteria:

  • High depth SNR (>7): Strong signal
  • Low odd-even mismatch: Consistent transit depth
  • Multiple transits observed: More reliable
  • Reasonable duration: Not too long or too short for orbit

Period Grid Sensitivity

The BLS periodogram is sensitive to period grid spacing. The autoperiod() method provides a conservative grid:

▼
python
复制代码
# Get automatic period grid periods = model.autoperiod(durations, minimum_period=1*u.day, maximum_period=10*u.day) print(f"Period grid has {len(periods)} points") # Use this grid with power() periodogram = model.power(periods, durations)

Tips:

  • Use autopower() for initial searches
  • Use finer grids around promising candidates
  • Period grid quality matters more for BLS than for Lomb-Scargle

Comparing BLS Results

To compare multiple peaks:

▼
python
复制代码
# Find top 5 peaks sorted_idx = np.argsort(periodogram.power)[::-1] top_5 = sorted_idx[:5] print("Top 5 candidates:") for i, idx in enumerate(top_5): period = periodogram.period[idx] power = periodogram.power[idx] duration = periodogram.duration[idx] stats = model.compute_stats(period, duration, periodogram.transit_time[idx]) print(f"\n{i+1}. Period: {period:.5f}") print(f" Power: {power:.2f}") print(f" Duration: {duration:.5f}") print(f" SNR: {stats['depth_snr']:.2f}") print(f" Transits: {stats['transit_count']}")

Phase-Folded Light Curve

After finding a candidate, phase-fold to visualize the transit:

▼
python
复制代码
# Fold the light curve at the best period phase = ((time.value - best_t0.value) % best_period.value) / best_period.value # Plot to verify transit shape import matplotlib.pyplot as plt plt.plot(phase, flux, '.') plt.xlabel('Phase') plt.ylabel('Flux') plt.show()

BLS vs Transit Least Squares (TLS)

Both methods search for transits, but differ in implementation:

Box Least Squares (BLS)

Pros:

  • Built into Astropy (no extra install)
  • Fast for targeted searches
  • Good statistical framework
  • compute_stats() provides detailed validation

Cons:

  • Simpler transit model (box shape)
  • Requires careful period grid setup
  • May be less sensitive to grazing transits

Transit Least Squares (TLS)

Pros:

  • More sophisticated transit models
  • Generally more sensitive
  • Better handles grazing transits
  • Automatic period grid is more robust

Cons:

  • Requires separate package
  • Slower for very long time series
  • Less control over transit shape

Recommendation: Try both! TLS is often more sensitive, but BLS is faster and built-in.

Integration with Preprocessing

BLS works best with preprocessed data. Consider this pipeline:

  1. Quality filtering: Remove flagged data points
  2. Outlier removal: Clean obvious artifacts
  3. Detrending: Remove stellar variability (rotation, trends)
  4. BLS search: Run period search on cleaned data
  5. Validation: Use compute_stats() to check candidate quality

Key Considerations

  • Preprocessing should preserve transit shapes (use gentle methods like flatten())
  • Don't over-process - too aggressive cleaning removes real signals
  • BLS needs reasonable period and duration ranges
  • Always validate with multiple metrics (power, SNR, odd-even)

Common Issues

Issue: No clear peak

Causes:

  • Transits too shallow
  • Wrong duration range
  • Period outside search range
  • Over-aggressive preprocessing

Solutions:

  • Try wider duration range
  • Extend period search range
  • Use less aggressive flatten() window
  • Check raw data for transits

Issue: Period is 2× or 0.5× expected

Causes:

  • Missing alternating transits
  • Data gaps
  • Period aliasing

Solutions:

  • Check both periods manually
  • Examine odd-even statistics
  • Look at phase-folded plots for both periods

Issue: High odd-even mismatch

Cause:

  • Not a planetary transit
  • Eclipsing binary
  • Instrumental artifact

Solution:

  • Check stats['depth_odd'] vs stats['depth_even']
  • May not be a transiting planet

Dependencies

▼
bash
复制代码
pip install astropy numpy matplotlib # Optional: lightkurve for preprocessing pip install lightkurve

References

Official Documentation

  • Astropy BLS Documentation
  • Astropy Time Series Guide

Key Papers

  • Kovács, Zucker, & Mazeh (2002): Original BLS paper - A&A 391, 369
  • Hartman & Bakos (2016): VARTOOLS implementation - A&C 17, 1

Related Resources

  • Lightkurve Tutorials
  • TLS GitHub - Alternative transit detection method

When to Use BLS

Use BLS when:

  • You want a fast, built-in solution
  • You need detailed validation statistics (compute_stats)
  • Working within the Astropy ecosystem
  • You want fine control over period grid

Use TLS when:

  • Maximum sensitivity is critical
  • Dealing with grazing or partial transits
  • Want automatic robust period grid
  • Prefer more sophisticated transit models

Use Lomb-Scargle when:

  • Searching for general periodic signals (not specifically transits)
  • Detecting stellar rotation, pulsation
  • Initial exploration of periodicity

For exoplanet detection, both BLS and TLS are valid choices. Try both and compare results!

研究

快捷安装

复制到终端去安装 Skill

npxskillsaddbenchflow-ai/skillsbench--skill"box-least-squares"
站长:程序员鱼皮 老鱼简历 商务合作 增值电信业务经营许可证
2026 鱼皮AI导航 | 沪ICP备19026706号-6
站长:程序员鱼皮 老鱼简历 商务合作 增值电信业务经营许可证
2026 鱼皮AI导航 | 沪ICP备19026706号-6
🔥热门工具
朱雀AI检测朱雀AI检测
朱雀AI检测助手朱雀AI检测助手
LovartLovart
老鱼简历老鱼简历
WritePass官网WritePass官网
面多多面多多
CursorCursor
星月写作星月写作
朱雀AI检测助手朱雀AI检测助手
零感AI零感AI
Skills 推荐
prime-numbers
面向图论中的素数问题求解,提供素性判定、整数分解、素数分布分析及模运算定理验证等核心能力,支持从基础试除到概率化与确定性算法的多层次策略选择,适用于理论推导与计算验证相结合的研究场景。
research
面向技术方案的深度调研能力,通过网络搜索、GitHub 仓库分析和多源证据采集,支持在实施前全面评估技术选型,确保每个推荐方案均基于官方文档、活跃代码库及真实应用案例等至少两项可靠依据。
rank-tracker
持续监控关键词在传统搜索结果与AI生成内容中的排名变化,分析趋势波动并识别显著升降,支持多维度对比竞争格局与SERP特征占有情况,输出可视化报告及可执行优化建议,适用于搜索引擎营销策略的动态调整。
research-grants
面向 NSF、NIH、DOE 和 DARPA 等联邦机构的科研基金申请,提供从选题定位、立项依据、技术路线设计、创新性阐述、可行性论证到预算编制与合规性审查的全流程支持,强调科学严谨性、跨学科整合能力及社会价值转化路径。
gsea-enrichment
执行基因集富集分析时,需先将通路数据库文件加载为字典格式,再进行富集计算,确保使用正确的数据结构与物种参数,支持多类功能注释数据库的整合分析与可视化。