arXiv subjects

Joohan Lee

Publications and source records attributed to Joohan Lee.

At least 19 recordsLinked to original sources

arXiv2024

Generative vs. Predictive Models in Massive MIMO Channel Prediction

Massive MIMO (mMIMO) systems are essential for 5G/6G networks to meet high throughput and reliability demands, with machine learning (ML)-based techniques, particularly autoencoders (AEs), showing promise for practical deployment. However, standard AEs struggle under noisy channel conditions, limiting their effectiveness. This work introduces a Vector Quantization-based generative AE model (VQ-VAE) for robust mMIMO cross-antenna channel prediction. We compare Generative and Predictive AE-based models, demonstrating that Generative models outperform Predictive ones, especially in noisy environments. The proposed VQ-VAE achieves up to 15 [dB] NMSE gains over standard AEs and about 9 [dB] over VAEs. Additionally, we present a complexity analysis of AE-based models alongside a diffusion model, highlighting the trade-off between accuracy and computational efficiency.

cs.IT↗

arXiv2024

Integrating Pre-Trained Language Model with Physical Layer Communications

The burgeoning field of on-device AI communication, where devices exchange information directly through embedded foundation models, such as language models (LMs), requires robust, efficient, and generalizable communication frameworks. However, integrating these frameworks with existing wireless systems and effectively managing noise and bit errors pose significant challenges. In this work, we introduce a practical ondevice AI communication framework, integrated with physical layer (PHY) communication functions, demonstrated through its performance on a link-level simulator. Our framework incorporates end-to-end training with channel noise to enhance resilience, incorporates vector quantized variational autoencoders (VQ-VAE) for efficient and robust communication, and utilizes pre-trained encoder-decoder transformers for improved generalization capabilities. Simulations, across various communication scenarios, reveal that our framework achieves a 50% reduction in transmission size while demonstrating substantial generalization ability and noise robustness under standardized 3GPP channel models.

cs.IT↗

arXiv2023

Software Entity Recognition with Noise-Robust Learning

Recognizing software entities such as library names from free-form text is essential to enable many software engineering (SE) technologies, such as traceability link recovery, automated documentation, and API recommendation. While many approaches have been proposed to address this problem, they suffer from small entity vocabularies or noisy training data, hindering their ability to recognize software entities mentioned in sophisticated narratives. To address this challenge, we leverage the Wikipedia taxonomy to develop a comprehensive entity lexicon with 79K unique software entities in 12 fine-grained types, as well as a large labeled dataset of over 1.7M sentences. Then, we propose self-regularization, a noise-robust learning approach, to the training of our software entity recognition (SER) model by accounting for many dropouts. Results show that models trained with self-regularization outperform both their vanilla counterparts and state-of-the-art approaches on our Wikipedia benchmark and two Stack Overflow benchmarks. We release our models, data, and code for future research.

cs.SE↗

arXiv2016

Supernova constraints on higher-dimensional cosmology with a phantom field

We use observational data on the magnitude-redshift relation for Type Ia supernovae together with constraints on the ages of the oldest stars to rule out a higher-dimensional extension of General Relativity with a negative kinetic-energy scalar field. This theory is of considerable physical interest because it produces accelerated expansion at both early and late times with a single new field, as in quintessential inflation scenarios. It is also of mathematical interest because it is characterized by an analytic expression for the macroscopic scale factor $a(t)$. We show that cosmological solutions of this theory can be usefully parametrized by a single quantity, the lookback time $τ_{\text{tr}}$ corresponding to the transition from deceleration to acceleration. Supernovae data from the recently released Supernova Cosmology Project Union 2.1 compilation single out a narrow range of values for $τ_{\text{tr}}$. In the context of the theory, however, these same values of $τ_{\text{tr}}$ imply that the universe is much older than the oldest observed stars.

gr-qc↗

arXiv2014

Cosmological Coincidence without Fine Tuning

We present a simple cosmological model in which a single, non-minimally coupled scalar field with a quartic potential is responsible for both inflation at early times and acceleration at late times. Little or no fine tuning is needed to explain why the present density of dark energy is comparable to that of pressureless matter. Dark energy is identified with the potential of the scalar field, which is sourced by the trace of the energy-momentum tensor. This becomes significant when matter has decoupled from radiation and become fully non-relativistic, so that $ϕ\proptoρ_m^{1/3}\proptoρ_{m,0}^{1/3}(a_0/a)\sim(10^{-120})^{1/3}(10^{10})\sim10^{-30}$ and $V\simϕ^4\sim10^{-120}$ in Planck units, as observed.

hep-th↗

arXiv2012

Dark spinor model with torsion and cosmology

We study cosmological consequences of the dark spinor model when torsion is included. Only some components of the torsion are allowed to be non-vanishing in homogeneous and isotropic cosmology, but there exist freedoms in the choice of these components which is consistent with the evolution equations. We exploit this and discuss several cases which can result in interesting cosmological consequences. Especially, we show that there exist exact cosmological solutions in which the Universe began its acceleration only recently and this solution is an attractor. This corresponds to a specific form of the torsion with a mild fine-tuning which can address the coincidence problem.

gr-qc↗

arXiv2012

Conformally-coupled dark spinor and FRW universe

We study conformal coupling of dark spinor fields to gravity and calculate the energy density and the pressure of the spinor in FRW spacetime. We consider the renormalizable potential of the spinor field. In the cases where the field is proportional to some power of the cosmic scale factor $a(t)$, we determine the Hubble parameter as a function of the scale factor and find analytic solutions for $a(t)$ when the spinor field matter dilutes as the universe expands. We discuss the possibility that both matter- and dark energy-dominated eras of our universe can be described by the dark spinor.

gr-qc↗

arXiv2012

Asymptotically Static Universe

We consider a cosmology in which the final stage of the Universe is neither accelerating nor decelerating, but approaches an asymptotic state where the scale factor becomes a constant value. In order to achieve this, we first bring in a scale factor with the desired property and then determine the details of the energy contents as a result of the cosmological evolution equations. We show that such a scenario can be realized if we introduce a generalized quintom model which consists of a scalar field and a phantom with a {\it negative} cosmological constant term. The standard cold dark matter with $w_m=0$ is also introduced. This is possible basically due to the balance between the matter and the {\it negative} cosmological constant which tend to attract and scalar field and phantom which repel in the asymptotic region. The stability analysis shows that this asymptotic solution is classically stable.

gr-qc↗

arXiv2010

Coupling of Brans-Dicke scalar field with Horava-Lifshitz gravity

We look for a Brans-Dicke type of generalization of the Horava-Lifshitz gravity. It is shown that such a generalization is possible within the detailed balance condition. Classically, the resulting theory reduces in the IR limit to the usual Brans-Dicke theory with a negative cosmological constant for certain values of parameters. We then consider homogeneous and isotropic cosmology and study the effects of the new terms appearing in the model.

hep-th↗

arXiv2010

Conformal Invariance in Einstein-Cartan-Weyl space

We consider conformally invariant form of the actions in Einstein, Weyl, Einstein-Cartan and Einstein-Cartan-Weyl space in general dimensions($>2$) and investigate the relations among them. In Weyl space, the observational consistency condition for the vector field determining non-metricity of the connection can be obtained from the equation of motion. In Einstein-Cartan space a similar role is played by the vector part of the torsion tensor. We consider the case where the trace part of the torsion is the Kalb-Ramond type of field. In this case, we express conformally invariant action in terms of two scalar fields of conformal weight -1, which can be cast into some interesting form. We discuss some applications of the result.

gr-qc↗

arXiv2009

De-Sitter nonlinear sigma model and accelerating universe

We consider a cosmology with a non-compact nonlinear sigma model.The target space is of de-Sitter type and four scalar fields are introduced.The potential is absent but cosmological constant term $Λ$ is added. One of the scalar fields is time dependent and the remaining three fields have no time dependence but only spatial dependence. We show that a very simple ansatz for the scalar fields results in the accelerating universe with an exponential expansion at late times. It is pointed out that the presence of the energy density and pressure coming from the spatial variation of the three scalar fields plays an essential role in our analysis which includes $Λ=0$ as a special case and it discriminate from the standard $Λ$-dominated acceleration. We perform a stability analysis of the solutions and find that some solutions are classically stable and attractor. We also present a non-perturbative solution which asymptotically approaches an exponential acceleration and discusspossible cosmological implications in relation with dark energy. It turns out that the equation of state approaches asymptotically $ω=-1$ both from above and below, but the crossing does not occur. It predicts present value of $ω\sim -1\mp 0.07$, which is within the region allowed by the oservational data. This solution also exhibits a power law expansion at early times, and the energy density of the scalar fields mimics that of the stiff matter.

gr-qc↗

arXiv2006

Supersymmetric Q-Lumps in the Grassmannian nonlinear sigma models

We construct the N=2 supersymmetric Grassmannian nonlinear sigma model for the massless case and extend it to massive N=2 model by adding an appropriate superpotential. We then study their BPS equations leading to supersymmetric Q-lumps carrying both topological and Noether charges. These solutions are shown to be always time dependent even sometimes involving multiple frequencies. Thus we illustrate explicitly that the time dependence is consistent with remaining supersymmetries of solitons.

hep-th↗

arXiv2006

Isospin particle on $S^{2}$ with arbitrary number of supersymmetries

We study the supersymmetric quantum mechanics of an isospin particle in the background of spherically symmetric Yang-Mills gauge field. We show that on $S^{2}$ the number of supersymmetries can be made arbitrarily large for a specific choice of the spherically symmetric SU(2) gauge field. However, the symmetry algebra containing the supercharges becomes nonlinear if the number of fermions is greater than two. We present the exact energy spectra and eigenfunctions, which can be written as the product of monopole harmonics and a certain isospin state. We also find that the supersymmetry is spontaneously broken if the number of supersymmetries is even.

hep-th↗

arXiv2005

A complete solution of a Constrained System: SUSY Monopole Quantum Mechanics

We solve the quantum mechanical problem of a charged particle on S^2 in the background of a magnetic monopole for both bosonic and supersymmetric cases by constructing Hilbert space and realizing the fundamental operators obeying complicated Dirac bracket relations in terms of differential operators. We find the complete energy eigenfunctions. Using the lowest energy eigenstates we count the number of degeneracies and examine the supersymmetric structure of the ground states in detail.

hep-th↗

arXiv2005

N=4 Supersymmetric Quantum Mechanics with Magnetic Monopole

We propose an N=4 supersymmetric quantum mechanics of a charged particle on a sphere in the background of Dirac magnetic monopole and study the system using the CP(1) model approach. We explicitly calculate the symmetry algebra taking the operator ordering ambiguity into consideration. We find that it is given by the superalgebra SU(1|2)x SU(2). We show that the Hamiltonian can be written in terms of the Casimir invariant of SU(2). Using this relation and the lower bound for angular momentm we obtain the energy spectrum. We then examine the ground energy sector to find that the N=4 supersymmetry is spontaneously broken to N=2 for certain values of the monopole charge.

hep-th↗

arXiv2005

Supersymmetric Monopole Quantum Mechanics on Sphere

We study N=2 supersymmetric quantum mechanics of a charged particle on sphere in the background of Dirac magnetic monopole. We adopt CP(1) model approach in which the monopole interaction is free of singularity. In order to exploit manifest U(1) covariance in the superspace formalism, we introduce a gauge covariant chiral superfield which is annihilated by the gauge covariant superderivative instead of the usual superderivative. We carry out the quantization of the resulting system and compute the quantum mechanical spectrum. We obtain the condition for the spontaneous breaking of supersymmetries explicitely in terms of the monopole charge and a parameter characterizing the operator ordering ambiguity. We find that the supersymmetry is spontaneously broken unless a certain combination of theses quantities satisfies some quantization condition.

hep-th↗

arXiv2003

Gauge symmetry enhancement in Hamiltonian formalism

We study the Hamiltonian structure of the gauge symmetry enhancement in the enlarged CP(N) model coupled with U(2) Chern-Simons term, which contains a free parameter governing explicit symmetry breaking and symmetry enhancement. After giving a general discussion of the geometry of constrained phase space suitable for the symmetry enhancement, we explicitly perform the Dirac analysis of our model and compute the Dirac brackets for the symmetry enhanced and broken cases. We also discuss some related issues.

hep-th↗

arXiv2003

Dirac Operators on Coset Spaces

The Dirac operator for a manifold Q, and its chirality operator when Q is even dimensional, have a central role in noncommutative geometry. We systematically develop the theory of this operator when Q=G/H, where G and H are compact connected Lie groups and G is simple. An elementary discussion of the differential geometric and bundle theoretic aspects of G/H, including its projective modules and complex, Kaehler and Riemannian structures, is presented for this purpose. An attractive feature of our approach is that it transparently shows obstructions to spin- and spin_c-structures. When a manifold is spin_c and not spin, U(1) gauge fields have to be introduced in a particular way to define spinors. Likewise, for manifolds like SU(3)/SO(3), which are not even spin_c, we show that SU(2) and higher rank gauge fields have to be introduced to define spinors. This result has potential consequences for string theories if such manifolds occur as D-branes. The spectra and eigenstates of the Dirac operator on spheres S^n=SO(n+1)/SO(n), invariant under SO(n+1), are explicitly found. Aspects of our work overlap with the earlier research of Cahen et al..

hep-th↗