86 research outputs found
Parametrized cosmological mass maps dataset
No full textParametrized cosmological mass maps dataset
This dataset consists of the non-tomographic training and testing set without noise and intrinsic alignments.
It was introduced in the following paper
Fluri, Janis, et al. "Cosmological constraints with deep learning from KiDS-450 weak lensing maps." Physical Review D 100.6 (2019): 063514.
Furthermore, this dataset is released with the following paper:
Perraudin, Nathanaël, et al. "Emulation of cosmological mass maps with conditional generative adversarial networks." arXiv preprint arXiv:2004.08139 (2020).
Code related to this dataset can be found in https://renkulab.io/projects/nathanael.perraudin/darkmattergan
Description
The simulation grid consists of different cosmologies assuming a flat LambdaCDM universe.
Each of these 57 configurations was run with different values of Omega_m and sigma_8, resulting in the following parameter grid.| Omega_m, sigma_8
0.101, 1.304
0.102, 1.125
0.103, 0.947
0.120, 1.178
0.123, 1.006
0.127, 0.836
0.137, 1.230
0.142, 1.063
0.148, 0.900
0.154, 1.281
0.156, 0.741
0.161, 1.119
0.169, 0.961
0.171, 1.331
0.178, 0.807
0.179, 1.173
0.188, 1.019
0.189, 0.659
0.196, 1.225
0.199, 0.870
0.207, 1.075
0.212, 0.727
0.219, 0.930
0.225, 1.129
0.227, 0.591
0.233, 0.791
0.238, 0.988
0.250, 0.658
0.254, 0.852
0.257, 1.043
0.269, 0.534
0.271, 0.723
0.273, 0.910
0.291, 0.601
0.291, 0.783
0.292, 0.966
0.311, 0.842
0.312, 0.664
0.314, 0.487
0.330, 0.898
0.332, 0.724
0.335, 0.552
0.352, 0.782
0.356, 0.614
0.370, 0.838
0.376, 0.673
0.382, 0.510
0.395, 0.730
0.402, 0.570
0.413, 0.784
0.421, 0.628
0.431, 0.475
0.440, 0.683
0.450, 0.533
0.458, 0.737
0.469, 0.589
0.487, 0.643
Each zip file in the dataset corresponds to 1 of these combinations and contains 12 files containing 1000 images.
The source galaxy redshift distribution corresponding to these maps is the full, non-tomographic redshift distribution n(z) from Fluri et. al.
The projected matter distribution was pixelised into images of size 128px x 128px, which correspond to 5deg x 5deg of the sky.
Eventually, the resulting dataset consists of 57 sets of 12'000 sky convergence maps for a total of samples.
Citations
If you use this dataset, please cite:
@article{perraudin2020emulation,
title={Emulation of cosmological mass maps with conditional generative adversarial networks},
author={Perraudin, Nathana{\"e}l and Marcon, Sandro and Lucchi, Aurelien and Kacprzak, Tomasz},
journal={arXiv preprint arXiv:2004.08139},
year={2020}
}
and
@article{fluri2019cosmological,
title={Cosmological constraints with deep learning from KiDS-450 weak lensing maps},
author={Fluri, Janis and Kacprzak, Tomasz and Lucchi, Aurelien and Refregier, Alexandre and Amara, Adam and Hofmann, Thomas and Schneider, Aurel},
journal={Physical Review D},
volume={100},
number={6},
pages={063514},
year={2019},
publisher={APS}
Double Momentum and Error Feedback for Clipping with Fast Rates and Differential Privacy
No full textStrong Differential Privacy (DP) and Optimization guarantees are two desirable properties for a method in Federated Learning (FL). However, existing algorithms do not achieve both properties at once: they either have optimal DP guarantees but rely on restrictive assumptions such as bounded gradients/bounded data heterogeneity, or they ensure strong optimization performance but lack DP guarantees. To address this gap in the literature, we propose and analyze a new method called Clip21-SGD2M based on a novel combination of clipping, heavy-ball momentum, and Error Feedback. In particular, for non-convex smooth distributed problems with clients having arbitrarily heterogeneous data, we prove that Clip21-SGD2M has optimal convergence rate and also near optimal (local-)DP neighborhood. Our numerical experiments on non-convex logistic regression and training of neural networks highlight the superiority of Clip21-SGD2M over baselines in terms of the optimization performance for a given DP-budget.Rustem Islamov and Aurelien Lucchi acknowledge the financial support of the Swiss National
Foundation, SNF grant No 207392. Peter Richtárik acknowledges the financial support of King
Abdullah University of Science and Technology (KAUST): i) KAUST Baseline Research Scheme,
ii) Center of Excellence for Generative AI, under award number 5940, iii) SDAIA-KAUST Center of Excellence in Artificial Intelligence and Data Science
Cosmological N-body simulations: a challenge for scalable generative models: Tensorflow checkpoints
No full text<p><strong>Tensorflow checkpoints: Cosmological N-body simulations: a challenge for scalable generative models</strong></p>
<p>This corresponds to the Tensorflow checkpoints for the experiments in the paper <strong>Cosmological N-body simulations: a challenge for scalable generative models</strong> by Nathanaël Perraudin, Ankit Srivastava, Aurelien Lucchi, Tomasz Kacprzak, Thomas Hofmann, Alexandre Refregier, Adam Amara.</p>
<pre><code>@inproceedings{perraudin2019cosmological,
title = {Cosmological N-body simulations: a challenge for scalable generative models},
author = {Nathana\"el, Perraudin and Ankit, Srivastava and Kacprzak, Tomasz and Lucchi, Aurelien and Hofmann, Thomas and R{\'e}fr{\'e}gier, Alexandre},
year = {2019},
archivePrefix = {arXiv},
eprint = {1908.05519},
url = {https://arxiv.org/abs/1908.05519},
}
</code></pre>
<p>Please check the assotiated github page <a href="https://github.com/nperraud/3DcosmoGAN">https://github.com/nperraud/3DcosmoGAN</a> for additional information.</p>
<p>This corresponds to the Tensorflow checkpoints for the experiments in the paper<br>
**Cosmological N-body simulations: a challenge for scalable generative models** by<br>
Nathanaël Perraudin, Ankit Srivastava, Aurelien Lucchi, Tomasz Kacprzak, Thomas Hofmann, Alexandre Refregier, Adam Amara.</p>
<p>Please check the assotiated github page <a href="https://github.com/nperraud/3DcosmoGAN">https://github.com/nperraud/3DcosmoGAN</a> for additional information.</p>
Signal propagation in transformers: theoretical perspectives and the role of rank collapse
No full textDigital media and intellectual property: Management of rights and consumer protection in a comparative analysis
No full textThe book provides a comparative and comprehensive analysis of the current technical, commercial and economical development in digital media. It describes the impact of new business and distribution models, the current legal and regulatory framework, social practices and consumer expectations associated with the use, distribution, and control of digital media products. In particular, the author analyzes the anti-circumvention provisions for technological protection measures and digital rights management systems enacted in the United States and in Europe, and their impact on consumer protection policy. The book concludes with an overview of the effects, and the possible solutions, under U.S. and EU law, posed by using contractual arrangements to expand intellectual property rights.This book provides a comparative and comprehensive analysis of the current state of technical, commercial and economical development in digital media. It describes the impact of new business and distribution models, the current legal and regulatory framework, social practices and consumer expectations associated with the use, distribution, and control of digital media products. In particular, the author analyzes the anti-circumvention provisions for technological protection measures and digital rights management systems enacted in the United States and in Europe, and their impact on consumer protection policy. The book concludes with an overview of the effects, and the possible solutions, under U.S. and EU law, posed by using contractual arrangements to expand intellectual property rights. © Springer-Verlag Berlin Heidelberg 2006. All rights are reserved
amyami187/effective_dimension: First release of the effective dimension code
No full textThis is the first release of the effective dimension code used to generate the graphs/analyses in the manuscript titled "The power of quantum neural networks" by Amira Abbas, David Sutter, Christa Zoufal, Aurelien Lucchi, Alessio Figalli and Stefan Woerner. This code is under the Apache 2.0 license
Cosmological N-body simulations: a challenge for scalable generative models: Tensorflow checkpoints
No full text<p>This corresponds to the Tensorflow checkpoints for the experiments in the paper<br>
**Cosmological N-body simulations: a challenge for scalable generative models** by<br>
Nathanaël Perraudin, Ankit Srivastava, Aurelien Lucchi, Tomasz Kacprzak, Thomas Hofmann, Alexandre Refregier, Adam Amara.</p>
<p>Please check the assotiated github page https://github.com/nperraud/3DcosmoGAN for additional information.</p>
Flash Scanning Electron Microscopy
No full textScanning Electron Microscopy (SEM) is an invaluable tool for biologists and neuroscientists to study brain structure at the intra- cellular level. While able to image tissue samples with up to 5nm isotropic resolution, image acquisition is prohibitively slow and limits the size of processed samples. In this work, we propose a novel approach to speeding up imaging when looking for specific structures. Unlike earlier methods, we explicitly balance the conflicting requirements of spending enough time scanning potential regions of interest to ensure that all targets are found while not wasting time on unpromising regions. This is achieved by using a Markov Random Field to model target locations and optimiz- ing scanning locations by using a Branch-and-Bound strategy. We show that our approach significantly outperforms state-of-the-art methods to locate mitochondria in brain tissue.CVLABCIM
Nbody 3D Histograms dataset
No full text
# This is the N-body simulations 3D images dataset used in the following paper:
* Scalable Generative Adversarial Networks for Multi-dimensional Images
Ankit Srivastava, Nathanaël Perraudin, Aurelien Lucchi, Tomasz Kacprzak, Thomas Hofmann, Alexandre Refregier, Adam Amara
The dataset does not contain the Nbody simulations as they have a very large size. Instead, we sliced the space into 256 x 256 x 256 cubical areas and counted the number of particules in each area. The result are 3D histograms, where the number of particles is a proxy for matter density.
Note that a the same Nbody simulation were used in this paper, but with a different way of building the histogram.
* Fast Cosmic Web Simulations with Generative Adversarial Networks
Andres C Rodriguez, Tomasz Kacprzak, Aurelien Lucchi, Adam Amara, Raphael Sgier, Janis Fluri, Thomas Hofmann, Alexandre Réfrégier
https://arxiv.org/abs/1801.09070v1
N-body simulation evolves a cosmological matter distribution over time, starting from soon after the big bang.
It represents matter density distribution as a finite set of massive particles, typically order of trillions.
The positions of these particles are modified due to gravitational forces and expansion of the cosmological volume due to cosmic acceleration.
N-body simulations use periodic boundary condition, where particles leaving the volume on one face enter it back from the opposite side.
## Short description of the data generation from Rordiguez et al. 2018:
We created N-body simulations of cosmic structures in boxes of size 100 Mpc and 500 Mpc with 512^3 and 1,024^3 particles respectively.
We used L-PICOLA [21] to create 10 and 30 independent simulation boxes for both box sizes.
The cosmological model used was ΛCDM (Cold Dark Matter) with Hubble constant H0 = 100, h = 70 km s−1 Mpc−1,
dark energy density Omega_Lambda = 0.72 and matter density Omega_m = 0.28.
We used the particle distribution at redshift z = 0.
For additional information, please check the README.md</p
- …
