arXiv subjects

Noa Kaplan

Publications and source records attributed to Noa Kaplan.

6 recordsLinked to original sources

arXiv2026

On Dominant Manifolds in Reservoir Computing Networks

Understanding how training shapes the geometry of recurrent network dynamics is a central problem in time-series modeling. We study the emergence of low-dimensional dominant manifolds in the training of Reservoir Computing (RC) networks for temporal forecasting tasks. For a general linear continuous-time reservoir in the infinite-data limit, we show that the training data generate an invariant subspace of the trained reservoir, whose dimension equals the number of dominant modes. We then specialize to a simplified diagonal linear reservoir, where we link the dominant eigenvalues and eigenvectors to the spectrum of a backward Dynamic Mode Decomposition matrix, which yields a finite-dimensional approximation of the backward-time Koopman operator of the system generating the training data. We illustrate the emergence of these dominant modes during training in simulation, and discuss how the analysis may be extended to nonlinear RC via tangent dynamics and differential p-dominance.

cs.LG↗

arXiv2025

Motion Generation: A Survey of Generative Approaches and Benchmarks

Motion generation, the task of synthesizing realistic motion sequences from various conditioning inputs, has become a central problem in computer vision, computer graphics, and robotics, with applications ranging from animation and virtual agents to human-robot interaction. As the field has rapidly progressed with the introduction of diverse modeling paradigms including GANs, autoencoders, autoregressive models, and diffusion-based techniques, each approach brings its own advantages and limitations. This growing diversity has created a need for a comprehensive and structured review that specifically examines recent developments from the perspective of the generative approach employed. In this survey, we provide an in-depth categorization of motion generation methods based on their underlying generative strategies. Our main focus is on papers published in top-tier venues since 2023, reflecting the most recent advancements in the field. In addition, we analyze architectural principles, conditioning mechanisms, and generation settings, and compile a detailed overview of the evaluation metrics and datasets used across the literature. Our objective is to enable clearer comparisons and identify open challenges, thereby offering a timely and foundational reference for researchers and practitioners navigating the rapidly evolving landscape of motion generation.

cs.CV↗

arXiv2020

Modelling light curves of bipolar core collapse supernovae from the equatorial plane

We use the two-components bipolar toy model of core collapse supernova (CCSN) ejecta to fit the rapid decline from maximum luminosity in the light curve of the type IIb CCSN SN 2018gk (ASASSN-18am). In this toy model we use a template light curve from a different CCSN that is similar to SN 2018gk, but that has no rapid drop in its light curve. The bipolar morphology that we model with a polar ejecta and an equatorial ejecta increases the maximum luminosity and causes a steeper decline for an equatorial observer, relative to a similar spherical explosion. The total energy and mass of our toy model for SN 2018gk are E=5e51 erg and M=2.7Mo. This explosion energy is more than what a neutrino-driven explosion mechanism can supply, implying that jets exploded SN 2018gk. These energetic jets likely shaped the ejecta to a bipolar morphology, as our toy model requires. We crudely estimate that f=2-5% of all CCSNe show this behavior, most being hydrogen deficient (stripped-envelope) CCSNe, as we observe them from the equatorial plane. We estimate the overall fraction of CCSNe that have a pronounced bipolar morphology to be 5-15% of all CCSNe.

astro-ph.HE↗

arXiv2020

Explaining recently studied intermediate luminosity optical transients (ILOTs) with jet powering

We apply the jet-powered ILOT scenario to two recently studied intermediate luminosity optical transients (ILOTs), and find the relevant shell mass and jets' energy that might account for the outbursts of these ILOTs. In the jet-powered ILOT scenario accretion disk around one of the stars of a binary system launches jets. The interaction of the jets with a previously ejected slow shell converts kinetic energy to thermal energy, part of which is radiated away. We apply two models of the jet-powered ILOT scenario. In the spherical shell model the jets accelerate a spherical shell, while in the cocoon toy model the jets penetrate into the shell and inflate hot bubbles, the cocoons. We find consistent results. For the ILOT (ILRT: intermediate luminosity red transient) SNhunt120 we find the shell mass and jets' energy to be Ms=0.5-1Mo and E2j=5e47erg, respectively. The jets' half opening angle is alpha=30-60 degrees. For the second peak of the ILOT (luminous red nova) AT2014ej we find these quantities to be Ms=1-2Mo and E2j=1.5e48erg, with alpha=20-30 degrees. The models cannot tell whether these ILOTs were powered by a stellar merger that leaves one star, or by mass transfer where both stars survived. In both cases the masses of the shells and energies of the jets suggest that the binary progenitor system was massive, with a combined mass of M1+M2>10Mo.

astro-ph.HE↗

arXiv2020

Jet-shaped geometrically modified light curves of core collapse supernovae

We build three simple bipolar ejecta models for core collapse supernovae (CCSNe), as expected when the explosion is driven by strong jets, and show that for an observer located in the equatorial plane of the ejecta, the light curve has a rapid luminosity decline, and even an abrupt drop. In calculating the geometrically modified photosphere we assume that the ejecta has an axisymmetrical structure composed of an equatorial ejecta and faster polar ejecta, and has a uniform effective temperature. At early times the photosphere in the polar ejecta grows faster than the equatorial one, leading to higher luminosity relative to a spherical explosion. The origin of the extra radiated energy is the jets. At later times the optical depth decreases faster in the polar ejecta, and the polar photosphere becomes hidden behind the equatorial ejecta for an observer in the equatorial plane, leading to a rapid luminosity decline. For a model where the jets inflate two low-density polar bubbles, the luminosity decline might be abrupt. This model enables us to fit the abrupt decline in the light curve of SN 2018don.

astro-ph.HE↗

arXiv2020

Emission peaks in the light curve of core collapse supernovae by late jets

We build a toy model where the central object, i.e., a newly born neutron star or a black hole, launches jets at late times and show that these jets might account for peaks in the light curve of some peculiar (i.e., having unusual light curves) core collapse supernovae (CCSNe) when the jets interact with the CCSN ejecta. We assume that the central object accretes fallback material and launches two short-lived opposite jets weeks to months after the explosion. We model each jet-ejecta interaction as a spherically symmetric `mini explosion' that takes place inside the ejecta. We assume that each `mini explosion' adds emission that is symmetric in time around the late peak, and with a rise in emission power that has the same slope as that of the main CCSN light curve. In total we use 12 parameters in the toy model. In our toy model late jets form stronger emission peaks than early jets. Late jets with a kinetic energy of only about one percent of the kinetic energy of the CCSN itself might form strong emission peaks. We apply our toy model to the brightest peak of the enigmatic CCSN iPTF14hls that has several extra peaks in its light curve. We can fit this emission peak with our toy model when we take the kinetic energy of the jets to be about one to two percent of the CCSN energy, and the shocked ejecta mass to be about three percent of the ejecta mass.

astro-ph.HE↗