arXiv subjects

Lina Guo

Publications and source records attributed to Lina Guo.

12 recordsLinked to original sources

arXiv2024

Orderings of the finite mixture with modified proportional hazard rate model

In this paper, we consider finite mixture models with modified proportional hazard rates. Sufficient conditions for the usual stochastic order and the hazard order are established under chain majorization. We study stochastic comparisons under different settings of T-transform for various values of chain majorization. We establish usual stochastic order and hazard rate order between two mixture random variables when a matrix of model parameters and mixing proportions changes to another matrix in some mathematical sense. Sufficient conditions for the star order and Lorenz order are established under weakly supermajorization. The results of this paper are illustrated with numerical examples.

math.ST↗

arXiv2024

Idempotence and Perceptual Image Compression

Idempotence is the stability of image codec to re-compression. At the first glance, it is unrelated to perceptual image compression. However, we find that theoretically: 1) Conditional generative model-based perceptual codec satisfies idempotence; 2) Unconditional generative model with idempotence constraint is equivalent to conditional generative codec. Based on this newfound equivalence, we propose a new paradigm of perceptual image codec by inverting unconditional generative model with idempotence constraints. Our codec is theoretically equivalent to conditional generative codec, and it does not require training new models. Instead, it only requires a pre-trained mean-square-error codec and unconditional generative model. Empirically, we show that our proposed approach outperforms state-of-the-art methods such as HiFiC and ILLM, in terms of Fr\'echet Inception Distance (FID). The source code is provided in https://github.com/tongdaxu/Idempotence-and-Perceptual-Image-Compression.

eess.IV↗

arXiv2023

Unified learning-based lossy and lossless JPEG recompression

JPEG is still the most widely used image compression algorithm. Most image compression algorithms only consider uncompressed original image, while ignoring a large number of already existing JPEG images. Recently, JPEG recompression approaches have been proposed to further reduce the size of JPEG files. However, those methods only consider JPEG lossless recompression, which is just a special case of the rate-distortion theorem. In this paper, we propose a unified lossly and lossless JPEG recompression framework, which consists of learned quantization table and Markovian hierarchical variational autoencoders. Experiments show that our method can achieve arbitrarily low distortion when the bitrate is close to the upper bound, namely the bitrate of the lossless compression model. To the best of our knowledge, this is the first learned method that bridges the gap between lossy and lossless recompression of JPEG images.

cs.CV↗

arXiv2023

Efficient Learned Lossless JPEG Recompression

JPEG is one of the most popular image compression methods. It is beneficial to compress those existing JPEG files without introducing additional distortion. In this paper, we propose a deep learning based method to further compress JPEG images losslessly. Specifically, we propose a Multi-Level Parallel Conditional Modeling (ML-PCM) architecture, which enables parallel decoding in different granularities. First, luma and chroma are processed independently to allow parallel coding. Second, we propose pipeline parallel context model (PPCM) and compressed checkerboard context model (CCCM) for the effective conditional modeling and efficient decoding within luma and chroma components. Our method has much lower latency while achieves better compression ratio compared with previous SOTA. After proper software optimization, we can obtain a good throughput of 57 FPS for 1080P images on NVIDIA T4 GPU. Furthermore, combined with quantization, our approach can also act as a lossy JPEG codec which has obvious advantage over SOTA lossy compression methods in high bit rate (bpp$>0.9$).

eess.IV↗

arXiv2022

Practical Learned Lossless JPEG Recompression with Multi-Level Cross-Channel Entropy Model in the DCT Domain

JPEG is a popular image compression method widely used by individuals, data center, cloud storage and network filesystems. However, most recent progress on image compression mainly focuses on uncompressed images while ignoring trillions of already-existing JPEG images. To compress these JPEG images adequately and restore them back to JPEG format losslessly when needed, we propose a deep learning based JPEG recompression method that operates on DCT domain and propose a Multi-Level Cross-Channel Entropy Model to compress the most informative Y component. Experiments show that our method achieves state-of-the-art performance compared with traditional JPEG recompression methods including Lepton, JPEG XL and CMIX. To the best of our knowledge, this is the first learned compression method that losslessly transcodes JPEG images to more storage-saving bitstreams.

eess.IV↗

arXiv2012

Stable surface solitons in truncated complex potentials

We show that surface solitons in the one-dimensional nonlinear Schrödinger equation with truncated complex periodic potential can be stabilized by linear homogeneous losses, which are necessary to balance gain in the near-surface channel arising from the imaginary part of potential. Such solitons become stable attractors when the strength of homogeneous losses acquires values from a limited interval and they exist in focusing and defocusing media. The domains of stability of surface solitons shrink with increase of the amplitude of imaginary part of complex potential.

physics.optics↗

arXiv2009

The two parameters (k, r) in the generalized statistics

Based on the generalized Boltzmann equation and the reverse function of the distribution function, we investigate the two-parameter generalized statistics and get an expression between the two parameters (k,r) and the physical quantities about the system considered. We find that the two parameters can define some characteristics of the system. As examples, this result can just return to the previous one obtained for Tsallis and statistics. For some complex systems, we may need the two-parameter statistics to describe.

cond-mat.stat-mech↗

arXiv2009

Heat capacity of the generalized two-atom and many-atom gas in nonextensive statistics

We have used the generalized two-atom ideal gas model in Tsallis statistics for the statistical description of a real gas. By comparing the heat capacity with the experimental results for the two-atom molecule gases such as N2, O2 and CO, we find that these gases appear extensive at normal temperature, but they may be nonextensive at the lower temperature. Furthermore, we study the heat capacity of the generalized many-atom gas model. We conclude that, for the many-atom gas with a high degree of freedom, a weak nonextensivity of 1-q<0 can lead to the instability.

cond-mat.stat-mech↗

arXiv2009

Exact boundary observability for nonautonomous quasilinear wave equations

By means of a direct and constructive method based on the theory of semiglobal $C^2$ solution, the local exact boundary observability is shown for nonautonomous 1-D quasilinear wave equations. The essential difference between nonautonomous wave equations and autonomous ones is also revealed.

math.OC↗

arXiv2008

Stability analysis of the classical ideal gas in nonextensive statistics and the negative specific heat

We present a stability analysis of the classical ideal gas in a new theory of nonextensive statistics and use the theory to understand the phenomena of negative specific heat in some self-gravitating systems. The stability analysis is made on the basis of the second variation of Tsallis entropy. It is shown that the system is thermodynamically unstable if the nonextensive parameter is q>5/3, which is exactly equivalent to the condition of appearance of the negative specific heat.

cond-mat.stat-mech↗

arXiv2008

Nonextensivity and q-distribution of a relativistic gas under an external electromagnetic field

We investigate the nonextensivity and the q-distribution of a relativistic gas under an external electromagnetic field. We derive a formula expression of the nonextensive parameter q based on the relativistic generalized Boltzmann equation, the relativistic q-H theorem and the relativistic version of q-power-law distribution function in the nonextensive q-kinetic theory. We thus provide the connection between the parameter 1-q and the differentiation of the temperature field of the gas as well as the four-potential with respect to time and space coordinates, and therefore present the nonextensivity for the gas a clearly physical meaning.

cond-mat.stat-mech↗

arXiv2006

The kappa parameter and kappa-distribution in kappa-deformed statistics for the systems in an external field

The kappa-deformed statistics has been studied in many papers. It is naturally important question for us to ask what should the kappa parameter stand for and under what physical situation should the kappa-deformed statistics be suitable for the statistical description of a system. In this paper, we have derived a formula expression of kappa parameter based on the kappa-H theorem, the kappa-velocity distribution and the generalized Boltzmann equation in the framework of kappa-deformed statistics. We thus obtain a physical interpretation for the kappa parameter different from zero with regard to the temperature gradient and the external force field. We show that, as the q-statistics based on Tsallis entropy, the kappa-deformed statistics may also be the candidate one suitable for the statistical description of the systems in external fields when being in the nonequilibrium stationary state, but has different physical characteristics. Namely, the kappa-distribution is found to describe the nonequilibrium stationary state of the system where the external force should be vertical to the temperature gradient.

cond-mat.stat-mech↗