arXiv subjects

Dimitrios Kosmopoulos

Publications and source records attributed to Dimitrios Kosmopoulos.

14 recordsLinked to original sources

arXiv2025

Dynamical Love Numbers for Black Holes and Beyond from Shell Effective Field Theory

We construct a novel effective field theory for a compact body coupled to gravity, whose key feature is that the dynamics of gravitational perturbations is explicitly determined by known solutions in black hole perturbation theory in four dimensions. In this way, the physics of gravitational perturbations in curved space are already encoded in the effective field theory, thus bypassing the need for the higher-order calculations that constitute a major hurdle in standard approaches. Concretely, we model the compact body as a spherical shell, whose finite size regulates short-distance divergences in four dimensions and whose tidal responses are described by higher-dimensional operators. As an application, we consider scalar perturbations and derive new results for scalar Love numbers through ${\cal O} (G^9)$ for Schwarzschild black holes and for generic compact bodies. Finally, our analysis reveals an intriguing structure of the scalar black-hole Love numbers in terms of the Riemann zeta function, which we conjecture to hold to all orders.

hep-th↗

arXiv2025

Observables and Unconstrained Spin Tensor Dynamics in General Relativity from Scattering Amplitudes

In a previous Letter, we showed that physical scattering observables for compact spinning objects in general relativity can depend on additional degrees of freedom in the spin tensor beyond those described by the spin vector alone. In this paper, we provide further details on the physics of these additional degrees of freedom, whose commutation relations and Poisson brackets are inherited from the underlying Lorentz symmetry, and on their consequence on observables. In particular, we give the waveform at leading order in Newton's constant and up to second order in the components of the spin tensor, and the conservative impulse, boost and spin kick, exhibiting spin magnitude change, through next-to-leading-order in Newton's constant and third order in the components of the spin tensor. We provide explicit examples -- a Newtonian two-particle bound system and a certain black-hole solution in an exotic matter-coupled gravitational theory -- that exhibit these degrees of freedom and are described by our four-dimensional and worldline field theories. We also discuss connections between these degrees of freedom and dynamical worldline multipole moments. We construct effective two-body Hamiltonians, we demonstrate explicitly that the extra degrees of freedom beyond the spin vector are necessary to describe the complete dynamics, and we explicitly remove certain unphysical singularities. Moreover, we show that the previously proposed eikonal (or radial action) formula correctly captures observables derived from the classical Hamiltonian. Finally, we comment on possible descriptions of the additional degrees of freedom from the perspective of Goldstone's theorem.

hep-th↗

arXiv2024

Conservative Spin Magnitude Change in Orbital Evolution in General Relativity

We show that physical scattering observables for compact spinning objects in general relativity can depend on additional degrees of freedom in the spin tensor beyond those described by the spin vector alone. The impulse, spin kick, and leading-order waveforms exhibit such a nontrivial dependence. A signal of this additional structure is the change in the magnitude of the spin vector under conservative Hamiltonian evolution, similar to our previous studies in electrodynamics. These additional degrees of freedom describe dynamical mass multipoles of compact objects and decouple for black holes. We also show that the conservative impulse, spin kick and change of the additional degrees of freedom are encoded in the eikonal phase.

hep-th↗

arXiv2023

Gravitational Self Force from Scattering Amplitudes in Curved Space

We employ scattering amplitudes in curved space to model the dynamics of a light probe particle with mass $m$ orbiting in the background spacetime induced by a heavy gravitational source with mass $M$. Observables are organized as an expansion in $m/M$ to all orders in $G$ -- the gravitational self-force expansion. An essential component of our analysis is the backreaction of the heavy source which we capture by including the associated light degrees of freedom. As illustration we consider a Schwarzschild background and verify geodesic motion as well as the first-order self-force correction to two-body scattering through ${\cal O}(G^3)$. Amplitudes in curved space offer several advantages, and further developments along these lines may advance the computation of gravitational-wave signals for extreme-mass-ratio inspirals.

hep-th↗

arXiv2023

Quantum Field Theory, Worldline Theory, and Spin Magnitude Change in Orbital Evolution

A previous paper~\cite{Bern:2022kto} identified a puzzle stemming from the amplitudes-based approach to spinning bodies in general relativity: additional Wilson coefficients appear compared to current worldline approaches to conservative dynamics of generic astrophysical objects, including neutron stars. In this paper we clarify the nature of analogous Wilson coefficients in the simpler theory of electrodynamics. We analyze the original field-theory construction, identifying definite-spin states some of which have negative norms, and relating the additional Wilson coefficients in the classical theory to transitions between different quantum spin states. We produce a new version of the theory which also has additional Wilson coefficients, but no negative-norm states. We match, through $\mathcal O(\alpha^2)$ and $\mathcal O(S^2)$, the Compton amplitudes of these field theories with those of a modified worldline theory with extra degrees of freedom introduced by releasing the spin supplementary condition. We build an effective two-body Hamiltonian that matches the impulse and spin kick of the modified field theory and of the worldline theory, displaying additional Wilson coefficients compared to standard worldline approaches. The results are then compactly expressed in terms of an eikonal formula. Our key conclusion is that, contrary to standard approaches, while the magnitude of the spin tensor is still conserved, the magnitude of the spin vector can change under conserved Hamiltonian dynamics and this change is governed by the additional Wilson coefficients. For specific values of Wilson coefficients the results are equivalent to those from a definite spin obeying the spin supplementary condition, but for generic values they are physically inequivalent. These results warrant detailed studies of the corresponding issues in general relativity.

hep-th↗

arXiv2022

Simplifying $D$-Dimensional Physical-State Sums in Gauge Theory and Gravity

We provide two independent systematic methods of performing $D$-dimensional physical-state sums in gauge theory and gravity in such a way so that spurious light-cone singularities are not introduced. A natural application is to generalized unitarity in the context of dimensional regularization or theories in higher spacetime dimensions. Other applications include squaring matrix elements to obtain cross sections, and decompositions in terms of gauge-invariant tensors.

hep-th↗

arXiv2022

Quantum Data Reduction with Application to Video Classification

We investigate a quantum video classification method using a hybrid algorithm. A quantum-classical step performs a data reduction on the video dataset and a quantum step -- which only has access to the reduced dataset -- classifies the video to one of k classes. We verify the method using sign videos and demonstrate that the reduced dataset contains enough information to successfully classify the data, using a quantum classification process. The proposed data reduction method showcases a way to alleviate the "data loading" problem of quantum computers for the problem of video classification. Data loading is a huge bottleneck, as there are no known efficient techniques to perform that task without sacrificing many of the benefits of quantum computing.

quant-ph↗

arXiv2022

Effective Field Theory Islands from Perturbative and Nonperturbative Four-Graviton Amplitudes

Theoretical data obtained from physically sensible field and string theory models suggest that gravitational Effective Field Theories (EFTs) live on islands that are tiny compared to current general bounds determined from unitarity, causality, crossing symmetry, and a good high-energy behavior. In this work, we present explicit perturbative and nonperturbative $2 \to 2$ graviton scattering amplitudes and their associated low-energy expansion in spacetime dimensions $D\geq 4$ to support this notion. Our new results include a first nonperturbative example consisting of a $D=4$, $\mathcal{N}=1$ supersymmetric field theory that is coupled weakly to gravity. We show that this nonperturbative model lies on the same islands identified using four-dimensional perturbative models based on string theory and minimally-coupled matter circulating a loop. Furthermore, we generalize the previous four-dimensional perturbative models based on string theory and minimally-coupled massive spin-0 and spin-1 states circulating in the loop to $D$ dimensions. Remarkably, we again find that the low-energy EFT coefficients lie on small islands. These results offer a useful guide towards constraining possible extensions of Einstein gravity.

hep-th↗

arXiv2022

Gravitational Effective Field Theory Islands, Low-Spin Dominance, and the Four-Graviton Amplitude

We analyze constraints from perturbative unitarity and crossing on the leading contributions of higher-dimension operators to the four-graviton amplitude in four spacetime dimensions, including constraints that follow from distinct helicity configurations. We focus on the leading-order effect due to exchange by massive degrees of freedom which makes the amplitudes of interest infrared finite. In particular, we place a bound on the coefficient of the $R^3$ operator that corrects the graviton three-point amplitude in terms of the $R^4$ coefficient. To test the constraints we obtain nontrivial effective field-theory data by computing and taking the large-mass expansion of the one-loop minimally-coupled four-graviton amplitude with massive particles up to spin 2 circulating in the loop. Remarkably, we observe that the leading EFT coefficients obtained from both string and one-loop field-theory amplitudes lie in small islands. The shape and location of the islands can be derived from the dispersive representation for the Wilson coefficients using crossing and assuming that the lowest-spin spectral densities are the largest. Our analysis suggests that the Wilson coefficients of weakly-coupled gravitational physical theories are much more constrained than indicated by bounds arising from dispersive considerations of $2 \to 2$ scattering. The one-loop four-graviton amplitudes used to obtain the EFT data are computed using modern amplitude methods, including generalized unitarity, supersymmetric decompositions and the double copy.

hep-th↗

arXiv2022

Binary Dynamics Through the Fifth Power of Spin at $\mathcal{O}(G^2)$

We use a previously developed scattering-amplitudes-based framework for determining two-body Hamiltonians for generic binary systems with arbitrary spin $S$. By construction this formalism bypasses difficulties with unphysical singularities or higher-time derivatives. This framework has been previously used to obtain the exact velocity dependence of the $\mathcal O(G^2)$ quadratic-in-spin two-body Hamiltonian. We first evaluate the $S^3$ scattering angle and two-body Hamiltonian at this order in $G$, including not only all operators corresponding to the usual worldline operators, but also an additional set due to an interesting subtlety. We then evaluate $S^4$ and $S^5$ contributions at $\mathcal O(G^2)$ which we confirm by comparing against aligned-spin results. We conjecture that a certain shift symmetry together with a constraint on the high-energy growth of the scattering amplitude specify the Wilson coefficients for the Kerr black hole to all orders in the spin and confirm that they reproduce the previously-obtained results through $S^4$.

hep-th↗

arXiv2021

A cross-domain recommender system using deep coupled autoencoders

Long-standing data sparsity and cold-start constitute thorny and perplexing problems for the recommendation systems. Cross-domain recommendation as a domain adaptation framework has been utilized to efficiently address these challenging issues, by exploiting information from multiple domains. In this study, an item-level relevance cross-domain recommendation task is explored, where two related domains, that is, the source and the target domain contain common items without sharing sensitive information regarding the users' behavior, and thus avoiding the leak of user privacy. In light of this scenario, two novel coupled autoencoder-based deep learning methods are proposed for cross-domain recommendation. The first method aims to simultaneously learn a pair of autoencoders in order to reveal the intrinsic representations of the items in the source and target domains, along with a coupled mapping function to model the non-linear relationships between these representations, thus transferring beneficial information from the source to the target domain. The second method is derived based on a new joint regularized optimization problem, which employs two autoencoders to generate in a deep and non-linear manner the user and item-latent factors, while at the same time a data-driven function is learnt to map the item-latent factors across domains. Extensive numerical experiments on two publicly available benchmark datasets are conducted illustrating the superior performance of our proposed methods compared to several state-of-the-art cross-domain recommendation frameworks.

cs.IR↗

arXiv2021

Stochastic Transformer Networks with Linear Competing Units: Application to end-to-end SL Translation

Automating sign language translation (SLT) is a challenging real world application. Despite its societal importance, though, research progress in the field remains rather poor. Crucially, existing methods that yield viable performance necessitate the availability of laborious to obtain gloss sequence groundtruth. In this paper, we attenuate this need, by introducing an end-to-end SLT model that does not entail explicit use of glosses; the model only needs text groundtruth. This is in stark contrast to existing end-to-end models that use gloss sequence groundtruth, either in the form of a modality that is recognized at an intermediate model stage, or in the form of a parallel output process, jointly trained with the SLT model. Our approach constitutes a Transformer network with a novel type of layers that combines: (i) local winner-takes-all (LWTA) layers with stochastic winner sampling, instead of conventional ReLU layers, (ii) stochastic weights with posterior distributions estimated via variational inference, and (iii) a weight compression technique at inference time that exploits estimated posterior variance to perform massive, almost lossless compression. We demonstrate that our approach can reach the currently best reported BLEU-4 score on the PHOENIX 2014T benchmark, but without making use of glosses for model training, and with a memory footprint reduced by more than 70%.

cs.CL↗

arXiv2021

Quadratic-in-Spin Hamiltonian at $\mathcal{O}(G^2)$ from Scattering Amplitudes

We obtain the quadratic-in-spin terms of the conservative Hamiltonian describing the interactions of a binary of spinning bodies in General Relativity through $\mathcal{O}(G^2)$ and to all orders in velocity. Our calculation extends a recently-introduced framework based on scattering amplitudes and effective field theory to consider non-minimal coupling of the spinning objects to gravity. At the order that we consider, we establish the validity of the formula proposed in \cite{Bern:2020buy} that relates the impulse and spin kick in a scattering event to the eikonal phase.

hep-th↗

arXiv2021

Variational Bayesian Sequence-to-Sequence Networks for Memory-Efficient Sign Language Translation

Memory-efficient continuous Sign Language Translation is a significant challenge for the development of assisted technologies with real-time applicability for the deaf. In this work, we introduce a paradigm of designing recurrent deep networks whereby the output of the recurrent layer is derived from appropriate arguments from nonparametric statistics. A novel variational Bayesian sequence-to-sequence network architecture is proposed that consists of a) a full Gaussian posterior distribution for data-driven memory compression and b) a nonparametric Indian Buffet Process prior for regularization applied on the Gated Recurrent Unit non-gate weights. We dub our approach Stick-Breaking Recurrent network and show that it can achieve a substantial weight compression without diminishing modeling performance.

stat.ML↗