http://arxiv.org/abs/1105.6116
Constraints on fNL from WMAP 7-year data using a neural network classifier
Casaponsa, Bridges, Curto, Barreiro, Hobson, Martinez-Gonzalez
* What does the neural network (NN) classifier classify to get to fNL?
Uses "multi-class" NN to measure non-Gaussianity in CMB. First NN is trained on simulated non-Gaussian CMB maps with a range of known fNL values. Uses wavelet coefficients; both HealPix wavelet and spherical Mexican hat wavelet. NN classifier agrees well with chisq minimization results. The fNL values that they get are similar to their previous results. Advantage of NN: does not require inversion of large covariance matrix [??], and is considerably faster.
* The covariance matrix of power spectrum appears here too. Then how does NN deal with error estimates?
http://arxiv.org/abs/1105.6298
A foreground cleaned CMB map from non-Gaussianity measurement
Saha
* Does he clean the CMB map given a non-G measurement? Does that make sense? Do foregrounds contribute noise or false signal to non-G measurements?
At a resolution of 1 deg, minimize the non-Gaussian properties of the cleaned map, which arize due to diffuse foreground emission components from the MW. This is a new method. Use a simple kurtosis statistic, and perform linear combination of 5 frequency maps, where the cleaned map has minimum kurtosis. This method is validated by Monte-Carlo simulations. Results match the WMAP's ILC [??] map. Some advantage over the ILC method--it doesn't generate negative bias in angular power spectrum [why?]. CMB results are robust with this method.
* ...but what about any non-Gaussianity detections? Does this method change their results?
http://arxiv.org/abs/1105.6333
Cosmology with the Square Kilometer Array (SKA)
Rawlings
* It's about reionization epoch, etc. Let's read the list of sciences it can do.
SKA's main method is the use of neutral Hydrogen (HI) survey, with the use of large-and-smart receptors in the form of aperture arrays [?]. Several phases: SKA0, SKA1, SKA2. It is important to cross correlate HI with other wavebands such as optical, NIR, other EoR surveys. It can contribute through gravitational lensing and H_0 studies.
* Mmm, I wish it listed the science in the abstract a bit more. I guess it must have been about matter distribution.
http://arxiv.org/abs/1105.6336
Radio to infrared spectra of late-type galaxies with Planck and WMAP data
Peel, Dickenson, Davies, Clements, Beswick
* Ooh, can we get spectral info at these wavelengths from CMB satellites?
Use the Planck compact source catalog + WMAP and other archival measurements to get construct continuum spectra of 3 nearby dusty star-forming galaxies. Fir to synchrotron, free-free and thermal dust models--but find that all are consistent with steep spectrum [?] synchrotron emission, with significant free-free [this is Bremsstrahlung] emission, which agrees better with SF rate based on Bremsstrahlung. Limits on anomalous microwave emission is placed. NGC4945 show hints of anomaly in microwave emission around 30GHz.
* I don't understand the significance of steep spectrum synchrotron emission, or the connection of free-free emission to star formation.
http://arxiv.org/abs/1105.6005
Analysis of KATRIN data using Bayesian inference
Riis, Hannestad, Weinheimer
* I just like these guys, who seem to be keen on connecting neutrino experiments with cosmology.
KATRIN will measure neutrino mass, but the data needs to be analyzed wrt standard model extension parameters for non-standard couplings to e.g. sterile neutrinos. They find COSMOMC package for MCMC works pretty well, instead of their usual method of frequentist minimization method.
* ...and they like COSMOMC.
http://arxiv.org/abs/1011.5692
Self-similar and charged spheres in the free-streaming approximation
Barreto, Rosales
* What?
They evolve non-adiabatic charged spherical distributions of matter. Almost half of the total initial mass is radiated away. The transport mechanism determines the way electric charge is redistributed.
* Oh, I guess they're considering cosmology in a charge-unbalanced universe.
http://arxiv.org/abs/1103.2767
Fast generation of ensembles of cosmological N-body simulations via mode-resampling
Schneider, Cole, Frenk, Szapudi
* I think I've read the abstract before, and this is a updated version.
Presented: an algorithm for quickly generating multiple realizations of N-body simulations for e.g., cosmo parameter estimations from surveys of LSS. Algorithm: resample the large-scale Fourier modes in a periodic N-body simulation box that accounts for NL coupling between large and small scales. Then add new large-scale mode realizations--this method recovers the NL power spectrum to sub-percent accuracy on scales larger than half the Nyquist frequency of the simulation box. From 20 N-body simulations, they can obtain power spectrum covariance matrix estimate that matches the estimator generated from 5000 realizations, with <20% error in all matrix elements. This new algorithm requires ~8 times fewer simulations than usual. This indicates that they understand the physical process that give rise to the covariance in the matter power spectrum: i.e., that the large-scale Fourier modes modulate both (1) the degree of structure growth through variation in the local matter density, and (2) the spatial frequency of small-scale perturbations through large-scale displacements. This algorithm can be useful for noise modeling when constraining cosmo params from WL and for galaxy surveys, rescaling summary statistics of N-body simulations, and any applications where the influence of the Fourier modes larger than the simulation size must be accounted for.
* I understand this abstract much better with this re-reading. Writing about it in this blog helps a lot too.
Uses "multi-class" NN to measure non-Gaussianity in CMB. First NN is trained on simulated non-Gaussian CMB maps with a range of known fNL values. Uses wavelet coefficients; both HealPix wavelet and spherical Mexican hat wavelet. NN classifier agrees well with chisq minimization results. The fNL values that they get are similar to their previous results. Advantage of NN: does not require inversion of large covariance matrix [??], and is considerably faster.
* The covariance matrix of power spectrum appears here too. Then how does NN deal with error estimates?
http://arxiv.org/abs/1105.6298
A foreground cleaned CMB map from non-Gaussianity measurement
Saha
* Does he clean the CMB map given a non-G measurement? Does that make sense? Do foregrounds contribute noise or false signal to non-G measurements?
At a resolution of 1 deg, minimize the non-Gaussian properties of the cleaned map, which arize due to diffuse foreground emission components from the MW. This is a new method. Use a simple kurtosis statistic, and perform linear combination of 5 frequency maps, where the cleaned map has minimum kurtosis. This method is validated by Monte-Carlo simulations. Results match the WMAP's ILC [??] map. Some advantage over the ILC method--it doesn't generate negative bias in angular power spectrum [why?]. CMB results are robust with this method.
* ...but what about any non-Gaussianity detections? Does this method change their results?
http://arxiv.org/abs/1105.6333
Cosmology with the Square Kilometer Array (SKA)
Rawlings
* It's about reionization epoch, etc. Let's read the list of sciences it can do.
SKA's main method is the use of neutral Hydrogen (HI) survey, with the use of large-and-smart receptors in the form of aperture arrays [?]. Several phases: SKA0, SKA1, SKA2. It is important to cross correlate HI with other wavebands such as optical, NIR, other EoR surveys. It can contribute through gravitational lensing and H_0 studies.
* Mmm, I wish it listed the science in the abstract a bit more. I guess it must have been about matter distribution.
http://arxiv.org/abs/1105.6336
Radio to infrared spectra of late-type galaxies with Planck and WMAP data
Peel, Dickenson, Davies, Clements, Beswick
* Ooh, can we get spectral info at these wavelengths from CMB satellites?
Use the Planck compact source catalog + WMAP and other archival measurements to get construct continuum spectra of 3 nearby dusty star-forming galaxies. Fir to synchrotron, free-free and thermal dust models--but find that all are consistent with steep spectrum [?] synchrotron emission, with significant free-free [this is Bremsstrahlung] emission, which agrees better with SF rate based on Bremsstrahlung. Limits on anomalous microwave emission is placed. NGC4945 show hints of anomaly in microwave emission around 30GHz.
* I don't understand the significance of steep spectrum synchrotron emission, or the connection of free-free emission to star formation.
http://arxiv.org/abs/1105.6005
Analysis of KATRIN data using Bayesian inference
Riis, Hannestad, Weinheimer
* I just like these guys, who seem to be keen on connecting neutrino experiments with cosmology.
KATRIN will measure neutrino mass, but the data needs to be analyzed wrt standard model extension parameters for non-standard couplings to e.g. sterile neutrinos. They find COSMOMC package for MCMC works pretty well, instead of their usual method of frequentist minimization method.
* ...and they like COSMOMC.
http://arxiv.org/abs/1011.5692
Self-similar and charged spheres in the free-streaming approximation
Barreto, Rosales
* What?
They evolve non-adiabatic charged spherical distributions of matter. Almost half of the total initial mass is radiated away. The transport mechanism determines the way electric charge is redistributed.
* Oh, I guess they're considering cosmology in a charge-unbalanced universe.
http://arxiv.org/abs/1103.2767
Fast generation of ensembles of cosmological N-body simulations via mode-resampling
Schneider, Cole, Frenk, Szapudi
* I think I've read the abstract before, and this is a updated version.
Presented: an algorithm for quickly generating multiple realizations of N-body simulations for e.g., cosmo parameter estimations from surveys of LSS. Algorithm: resample the large-scale Fourier modes in a periodic N-body simulation box that accounts for NL coupling between large and small scales. Then add new large-scale mode realizations--this method recovers the NL power spectrum to sub-percent accuracy on scales larger than half the Nyquist frequency of the simulation box. From 20 N-body simulations, they can obtain power spectrum covariance matrix estimate that matches the estimator generated from 5000 realizations, with <20% error in all matrix elements. This new algorithm requires ~8 times fewer simulations than usual. This indicates that they understand the physical process that give rise to the covariance in the matter power spectrum: i.e., that the large-scale Fourier modes modulate both (1) the degree of structure growth through variation in the local matter density, and (2) the spatial frequency of small-scale perturbations through large-scale displacements. This algorithm can be useful for noise modeling when constraining cosmo params from WL and for galaxy surveys, rescaling summary statistics of N-body simulations, and any applications where the influence of the Fourier modes larger than the simulation size must be accounted for.
* I understand this abstract much better with this re-reading. Writing about it in this blog helps a lot too.
No comments:
Post a Comment