اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدInflation might be caused by the right
72
0
0.0
(
0
)
تأليف
Gabriela Barenboim
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
We show that the scalar field that drives inflation can have a dynamical origin, being a strongly coupled right handed neutrino condensate. The resulting model is phenomenologically tightly constrained, and can be experimentally (dis)probed in the near future. The mass of the right handed neutrino obtained this way (a crucial ingredient to obtain the right light neutrino spectrum within the see-saw mechanism in a complete three generation framework) is related to that of the inflaton and both completely determine the inflation features that can be tested by current and planned experiments.
قيم البحث
اقرأ أيضاً
Inflationary models with a superheavy scale F-term hybrid inflation followed by an intermediate scale modular inflation are considered. The restrictions on the power spectrum P_R of curvature perturbation and the spectral index n_s from the recent da
ta within the power-law cosmological model with cold dark matter and a cosmological constant can be met provided that the number of e-foldings N_HI* suffered by the pivot scale k_*=0.002/Mpc during hybrid inflation is suitably restricted. The additional e-foldings needed for solving the horizon and flatness problems are generated by modular inflation with a string axion as inflaton. For central values of P_R and n_s, the grand unification scale comes out, in the case of standard hybrid inflation, close to its supersymmetric value M_GUT=2.86 x 10^16 GeV, the relevant coupling constant is relatively large (0.005-0.14), and N_HI* is between 10 and 21.7. In the shifted [smooth] hybrid inflation case, the grand unification scale can be identified with M_GUT for N_HI*=21 [N_HI*=18].
A simple extension of the minimal left-right symmetric supersymmetric grand unified theory model is constructed by adding two pairs of superfields. This naturally violates the partial Yukawa unification predicted by the minimal model. After including
supergravity corrections, we find that this extended model naturally supports hilltop F-term hybrid inflation along its trivial inflationary path with only a very mild tuning of the initial conditions. With a convenient choice of signs of the terms in the Kahler potential, we can reconcile the inflationary scale with the supersymmetric grand unified theory scale. All the current data on the inflationary observables are readily reproduced. Inflation is followed by non-thermal leptogenesis via the decay of the right-handed neutrinos emerging from the decay of the inflaton and any possible washout of the lepton asymmetry is avoided thanks to the violation of partial Yukawa unification. The extra superfields also assist us in reducing the reheat temperature so as to satisfy the gravitino constraint. The observed baryon asymmetry of the universe is naturally reproduced consistently with the neutrino oscillation parameters.
In this paper we introduce a provably stable architecture for Neural Ordinary Differential Equations (ODEs) which achieves non-trivial adversarial robustness under white-box adversarial attacks even when the network is trained naturally. For most exi
sting defense methods withstanding strong white-box attacks, to improve robustness of neural networks, they need to be trained adversarially, hence have to strike a trade-off between natural accuracy and adversarial robustness. Inspired by dynamical system theory, we design a stabilized neural ODE network named SONet whose ODE blocks are skew-symmetric and proved to be input-output stable. With natural training, SONet can achieve comparable robustness with the state-of-the-art adversarial defense methods, without sacrificing natural accuracy. Even replacing only the first layer of a ResNet by such a ODE block can exhibit further improvement in robustness, e.g., under PGD-20 ($ell_infty=0.031$) attack on CIFAR-10 dataset, it achieves 91.57% and natural accuracy and 62.35% robust accuracy, while a counterpart architecture of ResNet trained with TRADES achieves natural and robust accuracy 76.29% and 45.24%, respectively. To understand possible reasons behind this surprisingly good result, we further explore the possible mechanism underlying such an adversarial robustness. We show that the adaptive stepsize numerical ODE solver, DOPRI5, has a gradient masking effect that fails the PGD attacks which are sensitive to gradient information of training loss; on the other hand, it cannot fool the CW attack of robust gradients and the SPSA attack that is gradient-free. This provides a new explanation that the adversarial robustness of ODE-based networks mainly comes from the obfuscated gradients in numerical ODE solvers.
Topical phenomena in high-energy physics related to collision experiments of heavy nuclei (Little Bang) and early universe cosmology (Big Bang) involve far-from-equilibrium dynamics described by quantum field theory. One example concerns the role of
plasma instabilities for the process of thermalization in heavy-ion collisions. The reheating of the early universe after inflation may exhibit rather similar phenomena following a tachyonic or parametric resonance instability. Certain universal aspects associated to nonthermal fixed points even quantitatively agree, and considering these phenomena from a common perspective can be fruitful.
We show a hardness result for random smoothing to achieve certified adversarial robustness against attacks in the $ell_p$ ball of radius $epsilon$ when $p>2$. Although random smoothing has been well understood for the $ell_2$ case using the Gaussian
distribution, much remains unknown concerning the existence of a noise distribution that works for the case of $p>2$. This has been posed as an open problem by Cohen et al. (2019) and includes many significant paradigms such as the $ell_infty$ threat model. In this work, we show that any noise distribution $mathcal{D}$ over $mathbb{R}^d$ that provides $ell_p$ robustness for all base classifiers with $p>2$ must satisfy $mathbb{E}eta_i^2=Omega(d^{1-2/p}epsilon^2(1-delta)/delta^2)$ for 99% of the features (pixels) of vector $etasimmathcal{D}$, where $epsilon$ is the robust radius and $delta$ is the score gap between the highest-scored class and the runner-up. Therefore, for high-dimensional images with pixel values bounded in $[0,255]$, the required noise will eventually dominate the useful information in the images, leading to trivial smoothed classifiers.
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
