Exploring universe with neutral hydrogen + machine learningHayato Shimabukuro
This document discusses using machine learning to analyze signals from neutral hydrogen in the early universe. It provides background on the 21cm line signal from neutral hydrogen that can be used to study the epoch of reionization. It then discusses two ways machine learning has been used: (1) to estimate cosmological parameters from 21cm power spectra, and (2) to recover the distribution of ionized bubble sizes from 21cm power spectra. Artificial neural networks have been shown to accurately estimate parameters and reconstruct bubble size distributions from 21cm signals.
Exploring universe with neutral hydrogen + machine learningHayato Shimabukuro
This document discusses using machine learning to analyze signals from neutral hydrogen in the early universe. It provides background on the 21cm line signal from neutral hydrogen that can be used to study the epoch of reionization. It then discusses two ways machine learning has been used: (1) to estimate cosmological parameters from 21cm power spectra, and (2) to recover the distribution of ionized bubble sizes from 21cm power spectra. Artificial neural networks have been shown to accurately estimate parameters and reconstruct bubble size distributions from 21cm signals.
This document discusses using machine learning to analyze 21cm cosmology data from the Epoch of Reionization (EoR). It begins with background on the EoR and 21cm line signal. Current/future radio interferometers aim to detect the 21cm power spectrum to statistically map neutral hydrogen during the EoR. Machine learning techniques like artificial neural networks can be used as emulators to rapidly estimate EoR parameters from 21cm power spectra or recover ionized bubble size distributions that provide insights into the EoR. The document demonstrates how neural networks accurately recover input EoR parameters and bubble size distributions from simulated 21cm power spectrum data, highlighting their potential for 21cm cosmology analyses.
Multispectral remote sensors such as the Landsat Thematic Mapper and SPOT XS produce
images with a few relatively broad wavelength bands. Hyperspectral remote sensors, on the
other hand, collect image data simultaneously in dozens or hundreds of narrow, adjacent
spectral bands. These measurements make it possible to derive a continuous spectrum for each
image cell, as shown in the illustration below. After adjustments for sensor, atmospheric, and
terrain effects are applied, these image spectra can be compared with field or laboratory
reflectance spectra in order to recognize and map surface materials such as particular types of
vegetation or diagnostic minerals associated with ore deposits.
2022/3/24に開催した「オンプレML基盤 on Kubernetes」の資料です。機械学習モデルの開発者が、よりモデルの開発にのみ集中できるようにすることを目指して開発している「LakeTahoe(レイクタホ)」について紹介します。
https://ml-kubernetes.connpass.com/event/239859/
This document discusses using machine learning to analyze 21cm cosmology data from the Epoch of Reionization (EoR). It begins with background on the EoR and 21cm line signal. Current/future radio interferometers aim to detect the 21cm power spectrum to statistically map neutral hydrogen during the EoR. Machine learning techniques like artificial neural networks can be used as emulators to rapidly estimate EoR parameters from 21cm power spectra or recover ionized bubble size distributions that provide insights into the EoR. The document demonstrates how neural networks accurately recover input EoR parameters and bubble size distributions from simulated 21cm power spectrum data, highlighting their potential for 21cm cosmology analyses.
Multispectral remote sensors such as the Landsat Thematic Mapper and SPOT XS produce
images with a few relatively broad wavelength bands. Hyperspectral remote sensors, on the
other hand, collect image data simultaneously in dozens or hundreds of narrow, adjacent
spectral bands. These measurements make it possible to derive a continuous spectrum for each
image cell, as shown in the illustration below. After adjustments for sensor, atmospheric, and
terrain effects are applied, these image spectra can be compared with field or laboratory
reflectance spectra in order to recognize and map surface materials such as particular types of
vegetation or diagnostic minerals associated with ore deposits.
2022/3/24に開催した「オンプレML基盤 on Kubernetes」の資料です。機械学習モデルの開発者が、よりモデルの開発にのみ集中できるようにすることを目指して開発している「LakeTahoe(レイクタホ)」について紹介します。
https://ml-kubernetes.connpass.com/event/239859/
The document discusses the application of machine learning techniques to 21cm cosmology studies. It describes how artificial neural networks (ANNs) can be used as emulators to rapidly predict 21cm power spectra from cosmological parameters, bypassing the need for computationally expensive simulations. This allows ANNs to be combined with Markov chain Monte Carlo methods to efficiently estimate parameter posteriors. ANNs can also be applied to directly estimate parameters from 21cm power spectra or lightcones. The document outlines some open questions around fully characterizing uncertainties and obtaining rigorous posteriors when using ANN-based approaches in 21cm cosmology.
The document discusses using 21cm absorption lines, known as the 21cm forest, to probe properties of dark matter and the early universe. The 21cm forest signal is suppressed by both lower warm dark matter mass and higher X-ray heating. This is because both effects suppress the number of neutral hydrogen regions and masses of early halos. The document proposes using the 1D power spectrum of the 21cm forest to evaluate these effects and constrain dark matter scenarios. Future work could explore other dark matter models and primordial perturbations using this method.
Machine learning approaches are being applied in several ways to 21cm cosmology studies of the Epoch of Reionization (EoR):
1) Neural networks are used as emulators to rapidly predict 21cm power spectra from astrophysical parameters, speeding up Markov chain Monte Carlo parameter estimation.
2) Neural networks are trained on simulated 21cm power spectra to directly estimate astrophysical parameters, bypassing computationally expensive simulations.
3) Convolutional neural networks classify 21cm images to distinguish different sources driving reionization or compress image data into lower-dimensional summaries for likelihood-free inference of parameters.
Machine learning techniques are increasingly being used to tackle the statistical challenges of analyzing upcoming 21cm data and
1. Machine learning techniques can be applied to 21cm cosmology studies in various ways such as image reconstruction, signal detection, data analysis, simulation, and foreground subtraction.
2. Neural networks can be used to estimate cosmological parameters from 21cm power spectra or directly recover statistics like bubble size distributions from power spectra.
3. Studies have shown neural networks can accurately recover bubble size distributions from 21cm power spectra, even when including thermal noise at SKA sensitivity levels. This avoids information loss from incomplete image reconstruction.
4. Other work has used neural networks to reconstruct hydrogen distribution maps from galaxy surveys, demonstrating the potential of machine learning to connect 21cm signals to astrophysical sources and properties.
This document discusses the search for life in the universe. It begins by exploring definitions of life and the basic requirements for life as we understand it, such as the ability to grow, reproduce, and evolve. It then examines evidence that suggests life may have originated on Earth either from organic materials in space delivered by meteorites or around hydrothermal vents in the ocean. The document outlines NASA's roadmap for searching for life in our solar system and on exoplanets, including exploring moons of Jupiter and Saturn. It discusses using the Drake Equation to estimate the potential number of civilizations in our galaxy and the SETI Institute's plans to search for radio signals from extraterrestrial life using the upcoming SKA telescope.
1. In 1995, Michel Mayor and Didier Queloz directly discovered the first exoplanet orbiting the star 51 Pegasi b, for which they were later awarded the Nobel Prize in 2019.
2. There are now over 4000 known exoplanets that have been discovered, showing that our solar system is not unique in having planets.
3. The three main methods for detecting exoplanets are: measuring the Doppler effect of stars, detecting planetary transits that dim starlight, and directly imaging exoplanets.
- The document discusses the conditions needed for life to exist, namely liquid water, and the concept of the "habitable zone" where liquid water can exist on a planet based on its distance from its star.
- It notes that over 4,000 exoplanets have been discovered so far, and finding exoplanets within the habitable zones of their stars is key to searching for life in the universe. Approximately 50 exoplanets located in habitable zones have been discovered.
- The document also provides context on the timeline of life in the universe, noting that based on a "Cosmic Calendar" where the history of the universe is compressed into one year, life first emerged on September 21st, and human history only began in
1) The document discusses the early universe and the formation of light elements like hydrogen and helium in the first 3 minutes after the Big Bang through nuclear fusion processes.
2) It describes how the cosmic microwave background radiation provides evidence for the Big Bang theory, and how satellites like COBE and Planck have precisely measured tiny fluctuations in the CMB temperature that reveal information about the early universe.
3) Density fluctuations in the early universe seeded by cosmic inflation grew to form the large scale structures we observe today like dark matter halos, stars, and galaxies.
The document discusses the fate and evolution of the universe based on its density. It explains that if the density is greater than the critical density, the universe will collapse in a "Big Crunch", while if it is less it will expand forever. Observations find the density is approximately equal to the critical density, implying a flat universe. However, ordinary and dark matter only account for about 30% of the density, so some unknown "dark energy" is proposed to make up the remaining 70% and explain the observed accelerated expansion of the universe.
The document discusses key concepts in cosmology, including:
1) The universe appears homogeneous and isotropic at large scales, with only small fluctuations in density.
2) Olbers' paradox explains why the night sky is dark if the universe is assumed to be infinite and unchanging, which may be incorrect assumptions.
3) The expanding universe model provides a solution to Olbers' paradox, as the universe had a hot, dense beginning and a finite age, limiting the number of observable stars.
4) The fate of the universe depends on its total density - if density exceeds a critical threshold, the universe will eventually collapse, while a lower density means continued expansion.
1) Other galaxies, like the Milky Way, contain dark matter as evidenced by their rotation curves. Around 90% of the matter in the universe is dark matter.
2) While dark matter does not interact with electromagnetic waves, it can be observed through its gravitational effects such as gravitational lensing and its influence on galaxy cluster collisions.
3) Dark matter plays a key role in structure formation in the universe, starting with dark matter particles clumping together to form dark matter halos that then attract gas to form stars and galaxies.
1. The document discusses different types of galaxies including spiral, elliptical, irregular, and barred spiral galaxies.
2. It describes how Hubble classified galaxies into these categories and notes that galaxies do not evolve along Hubble's sequence.
3. Methods for measuring the distance to galaxies are explained, including using the Tully-Fisher relation to estimate the luminosity and distance of spiral galaxies based on measuring their circular velocity.
The document discusses key discoveries and concepts in astronomy:
1. Hubble discovered the expansion of the universe by measuring galaxies' receding velocities, which led to Hubble's law.
2. Some galaxies exhibit unusual activity and energy distributions, which is caused by violent events in their galactic nuclei, known as active galactic nuclei (AGN).
3. The central engine powering AGN is believed to be supermassive black holes that accrete material and produce powerful jets and radiation.
1. The document discusses the Milky Way Galaxy and other galaxies.
2. It describes how the existence of spiral galaxies was debated until Hubble observed variable stars in the Andromeda Nebula and measured its distance, finding it was outside the Milky Way Galaxy.
3. This supported Curtis's model that spiral nebulae exist outside the Milky Way and opened up the view that there are multiple galaxies in the universe beyond just the Milky Way.
The document discusses the Milky Way galaxy. It describes the Milky Way's structure, including the galactic disk, bulge, globular clusters, and halo. It explains how astronomers originally mapped the Milky Way by measuring the direction and distance to stars, using techniques like annual parallax and variable stars. The Milky Way is estimated to have formed around 13 billion years ago based on dating its oldest stars. The formation process likely involved the merging of smaller dwarf galaxies.
The document discusses neutron stars, pulsars, gamma-ray bursts, and black holes. It provides details on the properties of neutron stars, including their small size but large mass, extremely high density, rapid rotation, and strong magnetic fields. It also discusses the discovery of pulsars and how they provided evidence for rotating neutron stars. Gamma-ray bursts and fast radio bursts are also mentioned. The formation of black holes from objects with masses greater than the Schwarzschild radius is summarized.
1) Supernova explosions release energy equivalent to around 108-109 years of the sun's energy output.
2) There are two main types of supernovae - Type I with low hydrogen and Type II with lots of hydrogen. Their mechanisms differ but total energy output is similar.
3) Heavier elements up to iron form inside stars through fusion, while elements heavier than iron form via neutron capture processes during and after stellar evolution.
40. 21cm線と他波長観測
Kubota et al (2018)
•21cm線とライマンα輝線銀河の相互相関
Yoshiura et al (2018), Kubota et al (2018)
•21cm線と酸素[OIII]輝線銀河の相互相関
Moriwaki et al (2019)
•21cm線とCMBの相互相関
Yoshiura et al (2019)
21cm線と他波長は異なる系統誤差のため、21cm線シグナル検出に有効
•21cm線と背景X線の相互相関
Ma et al (2018)
Yoshiura et al (2019)
42. 21cm forest
•21cm forestで探る暗黒物質やインフ
レーション、原始ブラックホール
Shimabukuro et al (2014,2019,2020a), Villanueva &
Ichiki (2021), Kawasaki et al (2020)
宇宙再電離のみならず、宇宙論とも関係
•21cm forestで探る宇宙の加熱や銀河間物
質の温度状態
Furlanetto & Loeb (2002), Ciardi et al (2015),
Semelin (2015)
43. 21cm線観測への課題
Jelic et al 2008
21cm線は強烈な前景放射に埋もれている
前景放射除去?
or(and)
前景放射を避ける?
Santos 2005
~8 order
SKA日本再電離グループでは
21cm線と他波長の相互相関関数
で前景放射を軽減する方法を提案
Yoshiura et al (2018)
Kubota et al (2018)
55. Reconstruction of HI (ionized bubble) distribution
Fig. 5. Same as Figure 4, but for the LAEs z = 6.6. The large red open squares indicate the LAEs with spatially extended Lyα emission including Himiko
(Ouchi et al. 2009a) and CR7 (Sobral et al. 2015). See Shibuya et al. (2017b) for more details.
Input : LAE Output : HI distribution
Yoshiura +2021
Shimabukuro +2022
Input : 21cm PS
Output :ionized bubble size
distribution