How the AI Pipeline Works

The team, led by scientists from the SETI Institute and NASA’s Ames Research Center, deployed a convolutional neural network (CNN) trained on millions of simulated transit signals and known false positives. The network learned to distinguish the subtle, periodic dimming caused by a planet crossing its host star from instrumental artifacts, stellar variability, and cosmic rays.

একটি típico work‑flow এ siguiente ধাপগুলো অন্তর্ভুক্ত ছিল:

1. Raw光曲線からシステムティックノイズを除去するプレプロセッシング（コトレンドおよびピクセルレベルデコリレーション）
2. スライドウィンドウ方式で潜在的トランジットのような特徴を抽出
3. 各窓に対してCNNスコアを計算し、高確率の候補をランキング
4. 人間の専門家によるビジュアル検証と、既知のエクソプラネットカタログとのクロスマッチ

The result is a ranked list of 10,091 signals with a >90% probability of being genuine planetary transits, according to the team’s validation tests on injected synthetic planets.

From Candidates to Confirmed Worlds

Identifying a candidate is only the first step. Confirmation typically requires follow‑up observations—radial velocity measurements to gauge the planet’s mass, or additional transit photometry to refine the orbital period and depth. The researchers have already prioritized the top 100 candidates based on signal strength, host star brightness, and potential habitability.

বিশেষভাবে উত্সাহজনক হলো কিছু उम्मीदवार जो अपने माता‑तारा के आवासीय क्षेत्र में स्थित प्रतीत होते हैं, जिनके आकार पृथ्वी से सुपर‑पृथ्वी तक के रेंज में हैं। यदि इनमें से即便 कुछ भी सत्यापित हो जाए, तो हमारी आकाशगंगा में संभावित जीवन‑सहायक ग्रहों की संख्या में महत्वपूर्ण वृद्धि होगी।

To facilitate community vetting, the team has released the full catalog via the NASA Exoplanet Archive, complete with light‑curve cutouts, CNN confidence scores, and diagnostic plots. Amateur astronomers and professional teams alike can now download the data and attempt ground‑based confirmation using facilities such as the ESPRESSO spectrograph on the VLT or the upcoming NEID instrument on the WIYN telescope.

Implications for Astrobiology and Galactic Demographics

The sheer volume of candidates suggests that planetary systems are far more common than earlier estimates based on manual searches indicated. If even a modest fraction—say 10%—of these candidates prove to be bona fide planets, the Milky Way could host over a thousand new worlds from this single data release.

এই ডেটা সেট additionally promises insights into planetary formation mechanisms, especially for compact multi‑planet systems where gravitational interactions shape orbital architectures. By studying the period ratios and radius distributions of the candidates, astronomers can test theories of migration, resonant capture, and tidal damping.

Furthermore, the success of this AI‑driven approach paves the way for similar applications to upcoming missions such as PLATO (scheduled for launch in 2029) and the Roman Space Telescope, whose wide‑field imagers will generate petabytes of photometric data ripe for machine‑learning exploration.