FastICA as a LOFAR-EoR Foreground Cleaning Technique Filipe Abdalla and Emma Woodfield University College London with Saleem Zaroubi, Vibor Jelic, Panos.

Slides:



Advertisements
Similar presentations
Planck 2013 results, implications for cosmology
Advertisements

Suman Majumdar Department of Astronomy and Oskar Klein Centre Stockholm University Redshift Space Anisotropies in the EoR 21-cm Signal: what do they tell.
Combined Energy Spectra of Flux and Anisotropy Identifying Anisotropic Source Populations of Gamma-rays or Neutrinos Sheldon Campbell The Ohio State University.
05/03/2004 Measurement of Bunch Length Using Spectral Analysis of Incoherent Fluctuations Vadim Sajaev Advanced Photon Source Argonne National Laboratory.
Foreground cleaning in CMB experiments Carlo Baccigalupi, SISSA, Trieste.
Cleaned Three-Year WMAP CMB Map: Magnitude of the Quadrupole and Alignment of Large Scale Modes Chan-Gyung Park, Changbom Park (KIAS), J. Richard Gott.
EOR Detection Strategies Somnath Bharadwaj IIT Kharagpur.
Subtleties in Foreground Subtraction Adrian Liu, MIT mK 1 K 100 mK.
Indo – SA Joint Astronomy Workshop, August 2012 / 22 Study of Foregrounds and Limitations to EoR Detection Nithyanandan Thyagarajan N. Udaya Shankar Ravi.
Wed. 17th Sept Hamburg LOFAR Workshop.  Extract a cosmological signal from a datacube, the three axes of which are x and y positions, and frequency.
Interferometric Spectral Line Imaging Martin Zwaan (Chapters of synthesis imaging book)
Galaxy and Mass Power Spectra Shaun Cole ICC, University of Durham Main Contributors: Ariel Sanchez (Cordoba) Steve Wilkins (Cambridge) Imperial College.
Contamination of the CMB Planck data by galactic polarized emissions L. Fauvet, J.F. Macίas-Pérez.
Component Separation of Polarized Data Application to PLANCK Jonathan Aumont J-F. Macías-Pérez, M. Tristram, D. Santos
Cosmology with the 21 cm Transition Steve Furlanetto Yale University September 25, 2006 Steve Furlanetto Yale University September 25, 2006.
Independent Component Analysis (ICA) and Factor Analysis (FA)
A Quick Practical Guide to PCA and ICA Ted Brookings, UCSB Physics 11/13/06.
Laurent Itti: CS599 – Computational Architectures in Biological Vision, USC Lecture 7: Coding and Representation 1 Computational Architectures in.
(Rajib Saha,Pankaj Jain,Tarun Souradeep, astro-ph/ ) A Blind Estimation of Angular Power Spectrum of CMB Anisotropy from WMAP.
Matched Filter Search for Ionized Bubbles in 21-cm Maps Kanan K. Datta Dept. of Astronomy Stockholm University Oskar Klein Centre.
HELSINKI UNIVERSITY OF TECHNOLOGY LABORATORY OF COMPUTER AND INFORMATION SCIENCE NEURAL NETWORKS RESEACH CENTRE Variability of Independent Components.
Multidimensional Data Analysis : the Blind Source Separation problem. Outline : Blind Source Separation Linear mixture model Principal Component Analysis.
Separating Cosmological B-Modes with FastICA Stivoli F. Baccigalupi C. Maino D. Stompor R. Orsay – 15/09/2005.
The Statistical Properties of Large Scale Structure Alexander Szalay Department of Physics and Astronomy The Johns Hopkins University.
ElectroScience Lab IGARSS 2011 Vancouver Jul 26th, 2011 Chun-Sik Chae and Joel T. Johnson ElectroScience Laboratory Department of Electrical and Computer.
Wavelets, ridgelets, curvelets on the sphere and applications Y. Moudden, J.-L. Starck & P. Abrial Service d’Astrophysique CEA Saclay, France.
1 Patch Complexity, Finite Pixel Correlations and Optimal Denoising Anat Levin, Boaz Nadler, Fredo Durand and Bill Freeman Weizmann Institute, MIT CSAIL.
“First Light” From New Probes of the Dark Ages and Reionization Judd D. Bowman (Caltech) Hubble Fellows Symposium 2008.
Raman Research Institute, Bangalore, India Ravi Subrahmanyan (RRI, Bangalore) Ron Ekers & Aaron Chippendale (CAS) A Raghunathan & Nipanjana Patra (RRI,
Foreground subtraction or foreground avoidance? Adrian Liu, UC Berkeley.
130 cMpc ~ 1 o z~ = 7.3 Lidz et al ‘Inverse’ views of evolution of large scale structure during reionization Neutral intergalactic medium via HI.
COMMON EVALUATION FINAL PROJECT Vira Oleksyuk ECE 8110: Introduction to machine Learning and Pattern Recognition.
CMB & Foreground Polarisation CMB 2003 Workshop, Minneapolis Carlo Baccigalupi, SISSA/ISAS.
07/27/2004XFEL 2004 Measurement of Incoherent Radiation Fluctuations and Bunch Profile Recovery Vadim Sajaev Advanced Photon Source Argonne National Laboratory.
2010/12/11 Frequency Domain Blind Source Separation Based Noise Suppression to Hearing Aids (Part 1) Presenter: Cian-Bei Hong Advisor: Dr. Yeou-Jiunn Chen.
Joint analysis of Archeops and WMAP observations of the CMB G. Patanchon (University of British Columbia) for the Archeops collaboration.
Cosmic Microwave Background Carlo Baccigalupi, SISSA CMB lectures at TRR33, see the complete program at darkuniverse.uni-hd.de/view/Main/WinterSchoolLecture5.
SUNYAEV-ZELDOVICH EFFECT. OUTLINE  What is SZE  What Can we learn from SZE  SZE Cluster Surveys  Experimental Issues  SZ Surveys are coming: What.
PHY306 1 Modern cosmology 3: The Growth of Structure Growth of structure in an expanding universe The Jeans length Dark matter Large scale structure simulations.
Mário Santos1 EoR / 21cm simulations 4 th SKADS Workshop, Lisbon, 2-3 October 2008 Epoch of Reionization / 21cm simulations Mário Santos CENTRA - IST.
1 MaxEnt CNRS, Paris, France, July 8-13, 2006 “A Minimax Entropy Method for Blind Separation of Dependent Components in Astrophysical Images” Cesar.
Cosmic magnetism ( KSP of the SKA)‏ understand the origin and evolution of magnetism in the Galaxy, extragalactic objects, clusters and inter-galactic/-cluster.
NASSP Masters 5003F - Computational Astronomy Lecture 14 Reprise: dirty beam, dirty image. Sensitivity Wide-band imaging Weighting –Uniform vs Natural.
EBEx foregrounds and band optimization Carlo Baccigalupi, Radek Stompor.
1 The Planck view of CMB Contamination from Diffuse Foregrounds Carlo Baccigalupi On Behalf of the Planck Collaboration KITP Conference, April 2013.
Anisotropies in the gamma-ray sky Fiorenza Donato Torino University & INFN, Italy Workshop on High-Energy Messengers: connecting the non-thermal Extragalctic.
Foreground Contamination and the EoR Window Nithyanandan Thyagarajan N. Udaya Shankar Ravi Subrahmanyan (Raman Research Institute, Bangalore)
Blind Component Separation for Polarized Obseravations of the CMB Jonathan Aumont, Juan-Francisco Macias-Perez Rencontres de Moriond 2006 La.
WCRP Extremes Workshop Sept 2010 Detecting human influence on extreme daily temperature at regional scales Photo: F. Zwiers (Long-tailed Jaeger)
F Don Lincoln, Fermilab f Fermilab/Boeing Test Results for HiSTE-VI Don Lincoln Fermi National Accelerator Laboratory.
Introduction to Independent Component Analysis Math 285 project Fall 2015 Jingmei Lu Xixi Lu 12/10/2015.
An Introduction of Independent Component Analysis (ICA) Xiaoling Wang Jan. 28, 2003.
150GHz 100GHz 220GHz Galactic Latitude (Deg) A Millimeter Wave Galactic Plane Survey with the BICEP Polarimeter Evan Bierman (U.C. San Diego) and C. Darren.
Cosmic Microwave Background Carlo Baccigalupi, SISSA CMB lectures at TRR33, see the complete program at darkuniverse.uni-hd.de/view/Main/WinterSchoolLecture5.
The cross-correlation between CMB and 21-cm fluctuations during the epoch of reionization Hiroyuki Tashiro N. Aghanim (IAS, Paris-sud Univ.) M. Langer.
Planck working group 2.1 diffuse component separation review Paris november 2005.
The Planck view of CMB Contamination from Diffuse Foregrounds
EoR power spectrum systematics
Nicolas Fagnoni – Cosmology on Safari – 14th February 2017
EDGES: The ‘Global’ Perspective
Testing Primordial non-Gaussianities in CMB Anisotropies
Constraining the redshift of reionization using a “modest” array
Brain Electrophysiological Signal Processing: Preprocessing
Application of Independent Component Analysis (ICA) to Beam Diagnosis
HERA Imaging and Closure
Coherent analysis of CMB anisotropies CMB, structures and Foregrounds
Recovery of The Signal from the Epoch of Reionization
Apex scanning strategy & map-making
21 cm Foreground Subtraction in the Image Plane and in the uv-Plane
Separating E and B types of CMB polarization on an incomplete sky Wen Zhao Based on: WZ and D.Baskaran, Phys.Rev.D (2010) 2019/9/3.
Presentation transcript:

FastICA as a LOFAR-EoR Foreground Cleaning Technique Filipe Abdalla and Emma Woodfield University College London with Saleem Zaroubi, Vibor Jelic, Panos Labropoulos and the LOFAR-EoR working group

Problem Outline 21cm signal dominated both by foregrounds and by noise. Currently most foreground cleaning methods are parametric, e.g. polynomial. Non-parametric methods have emerged - Wp smoothing (Harker 09). Other, powerful techniques have been used on the CMB...

Problem Outline 21cm signal dominated both by foregrounds and by noise. Currently most foreground cleaning methods are parametric, e.g. polynomial. Non-parametric methods have emerged - Wp smoothing (Harker 09). Other, powerful techniques have been used on the CMB... Harker et al.2010

Problem Outline 21cm signal dominated both by foregrounds and by noise. Currently most foreground cleaning methods are parametric, e.g. polynomial. Non-parametric methods have emerged - Wp smoothing (Harker 09). Other, powerful techniques have been used on the CMB... Maino et al. 02

Independent Component Analysis and FastICA x = A s A is the mixing matrix. If we can find a matrix W such that s = W x we have effectively sorted a mixed signal into its individual components. “PDF of a mixture of independent components more Gaussian than the PDF of any individual component.” FastICA uses negentropy as a measure of Gaussianity. Many ICA methods exist (e.g. SMICA, GMCA, MCA), we have started with FastICA as easiest to implement. FastICA is an independent component analysis algorithm (Hyvärinen A., 1999, IEEE Trans. on Neural Networks, 10,626;Hyvärinen A., Karhunen J., Oja E., 2001, Independent Component Analysis. John Wiley and Sons) ICA methods are `blind’ and non-parametric. FastICA recovered the angular PS at the % level and could process all-sky maps in ~ 10 minutes. (Maino02)

Data Simulations: 21cm Signal 10°x10° data cube with 170 slice in frequency, separated by 0.5 Mhz. Frequency range of 115 – 200 MHz (z ~ 11.3 – 6.1) Box size of 1.8 Gpc over 512 pixels – a resolution of ~3MPc/pixel. 21cmFAST (Mesinger A., Furlanetto S., Cen R., 2011, MNRAS, 411,955) Minimum contributing halo temperature of 1e4 K. 21cm map at 150MHz for a 10°x10° observing window. Temperature scale in K.

Data Simulations: Foregrounds Simulations from Jelic V., Zaroubi S., Labropoulos P. et al. 2008, MNRAS, 389, 1319 Galactic synchrotron radiation, galactic free-free emission Extragalactic radiation from radio galaxies and clusters Foreground map at 150MHz for a 10°x10° observing window. Temperature scale in K.

Data Simulations: Noise Noise map at 150MHz for a 10°x10° observing window. Temperature scale in K. A measurement set of our simulation was filled with a Gaussian distribution of random numbers and a correlated noise map created with an imager. Each map was normalized to the expected rms noise of the LOFAR experiment as set out in Labropoulos et al. (2009); Jeli´c et al. (2008). e.g. 52mK at 150MHz for 600 hours of LOFAR observing time.

Data Simulations: Beam Forming Beam forming simulated by convolving with a gaussian PSF normalized to 3 arcmin beam width at 150MHz. OriginalAdjusted for convolution

Results: Residual variance recovery FastICA outputs “residuals” consisting of the reconstructed 21cm signal, noise and fitting errors. FastICA can be carried out in both fourier and real space with no difference between the methods. To estimate the reconstructed 21 variance we remove the noise by hand to check how much signal is mis- fitted.... Simulated residuals Reconstructed residuals

Results: 21cm variance recovery Excess variance is recovered, however it underestimates the simulated variance at almost all redshifts. This signal has been misfit – probably as a result of noise leakage into the foregrounds.

Simulated 21cm signal - beamformed Simulated 21cm map at 150MHz for a 10°x10° observing window. Signal has been adjusted for beam forming. Temperature scale in K.

Reconstructed 21cm signal - beamformed Reconstructed 21cm map at 150MHz for a 10°x10° observing window. Temperature scale in K. Large scale structure recovered but a lot of excess small scale structure – probably as a result of noise leakage into the foregrounds.

Results: 21cm Powerspectrum – 169 Mhz We look at the variance as a function of scale using the 2D angular power spectrum. We find the angular power spectrum at each frequency and then average over 5 slices (2.5 MHz). Error Bars:

Results: 21cm Powerspectrum – 149 MHz The small scale discrepancy could be an indication that FastICA is not robust to the noise in this method, or it could be a resolution effect.

Results: 21cm Powerspectrum – 129 MHz At the lower end of the frequency range the fit degrades significantly.

Conclusions ICA has been a successful method for cleaning CMB data. This is the first attempt at applying the FastICA method to LOFAR-EoR data. FastICA as foreground cleaning technique would be able to retrieve the EoR signal for a LOFAR set-up, and is especially robust for large scales. –This is not unsurprising as there is more noise at smaller scales and FastICA is not robust to noise. To do list: –We need to run FastICA on a higher resolution grid to test whether the poor small scale recovery is due to resolution effects or an inherent problem with the noise robustness of FastICA –Other methods have to be tested which can be more robust to noise and potentially recover spectra info, i.e. SMICA...