共找到 20 条结果
Kullback-Leibler (KL) divergence is a fundamental concept in information theory that quantifies the discrepancy between two probability distributions. In the context of Variational Autoencoders (VAEs), it serves as a central regularization term, imposing structure on the latent space and thereby enabling the model to exhibit generative capabilities. In this work, we present a detailed derivation of the closed-form expression for the KL divergence between Gaussian distributions, a case of particular importance in practical VAE implementations. Starting from the general definition for continuous random variables, we derive the expression for the univariate case and extend it to the multivariate setting under the assumption of diagonal covariance. Finally, we discuss the interpretation of each term in the resulting expression and its impact on the training dynamics of the model.
We provide a functional characterization of isometries between non-reversible Finsler manifolds, in the form of a generalization of the Myers-Nakai Theorem for Riemannian manifolds. We show that, since non-reversible Finsler manifolds are a fundamentally asymmetric object, such a result can not be obtained by means of a symmetric function space, and we define the appropriate asymmetric structure needed to describe all possible isometries between this class of manifolds. The result is based on the ideas used in a previous generalization for reversible Finsler manifolds proved in \cite{GJR-13}, in which the normed algebra of $C^1$-smooth Lipschitz functions is used. To reflect the quasi-metric structure of non-reversible Finsler manifolds, this normed algebra had to be modified to include the cone of smooth semi-Lipschitz functions, resulting in a partial loss of the normed space and algebra structures. In order to achieve the desired result, we define new algebraic/quasi-metric structures to model the behavior of the aforementioned function space.
Human spoken language has long been the subject of scientific investigation, particularly with regard to the mechanisms underpinning speech production. Likewise, the study of animal communications has a substantial literature, with many studies focusing on vocalisation. More recently, there has been growing interest in comparing animal communications and human speech. However, it is proposed here that such a comparison necessitates the appraisal of a minimum set of critical phenomena: i) the number of degrees-of-freedom of the vocal apparatus, ii) the ability to control those degrees-of-freedom independently, iii) the properties of the acoustic environment in which communication takes place, iv) the perceptual salience of the generated sounds, v) the degree to which sounds are contrastive, vi) the presence/absence of compositionality, and vii) the information rate(s) of the resulting communications.
A short phenomenological account of the genesis and evolution of the universe is presented with emphasis on the primordial phases as well as its physical composition, i.e. dark matter and dark energy. We discuss Einstein's theory of General Relativity and its consequences for the birth of modern relativistic astrophysics. We introduce the Big-Bang theory of Mons. Lemaitre as well as the competing theory of the Steady State Universe of Fred Hoyle. Since Big-Bang theory appeared quite in agreement with Christian doctrine of creation, Pope Pius XII delivered a message to the pontifical Academy of Sciences in 1951 claiming a certain agreement between the creation account in the book of Genesis and the Big-Bang theory (a concordist view), a position which he did not repeat later. On the other hand, Lemaitre always kept separate the scientific and theological planes as two parallel "lines" never intersecting, i.e., as two complementary "magisteria". Similar kind of tensions, between science and theology, emerge also today with the Hartle-Hawking solution to the Wheeler-DeWitt equation in quantum cosmology and its related speculations. To avoid some sort of confusion between theological a
At the last few years, multimedia communication has been developed and improved rapidly in order to enable users to communicate between each other over the internet. Generally, multimedia communication consists of audio and video communication. However, this research concentrates on audio conferencing only. The audio translation between protocols is a very critical issue, because it solves the communication problems between any two protocols. So, it enables people around the world to talk with each other even they use different protocols. In this research, a real time audio translation module between two protocols has been done. These two protocols are: InterAsterisk eXchange Protocol (IAX) and Real-Time Switching Control Protocol (RSW), which they are widely used to provide two ways audio transfer feature. The solution here is to provide inter-working between the two protocols which they have different media transports, audio codecs, header formats and different transport protocols for the audio transmission. This translation will help bridging the gap between the two protocols by providing inter-working capability between the two audio streams of IAX and RSW. Some related works h
Automated program analysis is a pivotal research domain in many areas of Computer Science -- Formal Methods and Artificial Intelligence, in particular. Due to the undecidability of the problem of program equivalence, comparing two programs is highly challenging. Typically, in order to compare two programs, a relation between both programs' sets of variables is required. Thus, mapping variables between two programs is useful for a panoply of tasks such as program equivalence, program analysis, program repair, and clone detection. In this work, we propose using graph neural networks (GNNs) to map the set of variables between two programs based on both programs' abstract syntax trees (ASTs). To demonstrate the strength of variable mappings, we present three use-cases of these mappings on the task of program repair to fix well-studied and recurrent bugs among novice programmers in introductory programming assignments (IPAs). Experimental results on a dataset of 4166 pairs of incorrect/correct programs show that our approach correctly maps 83% of the evaluation dataset. Moreover, our experiments show that the current state-of-the-art on program repair, greatly dependent on the programs'
Two of the fundamental no-go theorems of quantum information are the no-cloning theorem (that it is impossible to make copies of general quantum states) and the no-teleportation theorem (the prohibition on telegraphing, or sending quantum states over classical channels without pre-shared entanglement). They are known to be equivalent, in the sense that a collection of quantum states is telegraphable if and only if it is clonable. Our main result suggests that this is not the case when computational efficiency is considered. We give a collection of quantum states and quantum oracles relative to which these states are efficiently clonable but not efficiently telegraphable. Given that the opposite scenario is impossible (states that can be telegraphed can always trivially be cloned), this gives the most complete quantum oracle separation possible between these two important no-go properties. We additionally study the complexity class clonableQMA, a subset of QMA whose witnesses are efficiently clonable. As a consequence of our main result, we give a quantum oracle separation between clonableQMA and the class QCMA, whose witnesses are restricted to classical strings. We also propose a
In 2017, Igusa and Todorov gave a bijection between signed exceptional sequences and ordered partial clusters. In this paper, we show that every term in an exceptional sequence is either relatively projective or relatively injective or both and we refine this bijection to one between projectively signed exceptional sequences and ordered partial positive clusters. We also give a characterization of relatively projective/injective objects in terms of supports of the objects in the exceptional sequence.
What is the optimal way to deform a projective hypersurface into another one? In this paper we will answer this question adopting the point of view of measure theory, introducing the optimal transport problem between complex algebraic projective hypersurfaces. First, a natural topological embedding of the space of hypersurfaces of a given degree into the space of measures on the projective space is constructed. Then, the optimal transport problem between hypersurfaces is defined through a constrained dynamical formulation, minimizing the energy of absolutely continuous curves which lie on the image of this embedding. In this way an inner Wasserstein distance on the projective space of homogeneous polynomials is introduced. This distance is finer than the Fubini-Study one. The innner Wasserstein distance is complete and geodesic: geodesics corresponds to optimal deformations of one algebraic hypersurface into another one. Outside the discriminant this distance is induced by a smooth Riemannian metric, which is the real part of an explicit Hermitian structure. Moreover, this Hermitian structure is Kähler and the corresponding metric is of Weil-Petersson type. To prove these results w
We present a model that takes into account the coupling between evolutionary game dynamics and social influence. Importantly, social influence and game dynamics take place in different domains, which we model as different layers of a multiplex network. We show that the coupling between these dynamical processes can lead to cooperation in scenarios where the pure game dynamics predicts defection. In addition, we show that the structure of the network layers and the relation between them can further increase cooperation. Remarkably, if the layers are related in a certain way, the system can reach a polarized metastable state.These findings could explain the prevalence of polarization observed in many social dilemmas.
Compensating changes between a subjects' training and testing session in Brain Computer Interfacing (BCI) is challenging but of great importance for a robust BCI operation. We show that such changes are very similar between subjects, thus can be reliably estimated using data from other users and utilized to construct an invariant feature space. This novel approach to learning from other subjects aims to reduce the adverse effects of common non-stationarities, but does not transfer discriminative information. This is an important conceptual difference to standard multi-subject methods that e.g. improve the covariance matrix estimation by shrinking it towards the average of other users or construct a global feature space. These methods do not reduces the shift between training and test data and may produce poor results when subjects have very different signal characteristics. In this paper we compare our approach to two state-of-the-art multi-subject methods on toy data and two data sets of EEG recordings from subjects performing motor imagery. We show that it can not only achieve a significant increase in performance, but also that the extracted change patterns allow for a neurophys
In this paper, research on AI based modeling technique to optimize development of new alloys with necessitated improvements in properties and chemical mixture over existing alloys as per functional requirements of product is done. The current research work novels AI in lieu of predictions to establish association between material and product customary. Advanced computational simulation techniques like CFD, FEA interrogations are made viable to authenticate product dynamics in context to experimental investigations. Accordingly, the current research is focused towards binding relationships between material design and product design domains. The input to feed forward back propagation prediction network model constitutes of material design features. Parameters relevant to product design strategies are furnished as target outputs. The outcomes of ANN shows good sign of correlation between material and product design domains. The study enriches a new path to illustrate material factors at the time of new product development.
Radiative energy transfer between closely spaced bodies is known to be significantly larger than that predicted by classical radiative transfer because of tunneling due to evanescent waves. Theoretical analysis of near--field radiative transfer is mainly restricted to radiative transfer between two half--spaces or spheres treated in the dipole approximation (very small sphere) or proximity force approximation (radius of sphere much greater than the gap). Sphere--sphere or sphere--plane configurations beyond the dipole approximation or proximity force approximation have not been attempted. In this work, the radiative energy transfer between two adjacent non--overlapping spheres of arbitrary diameters and gaps is analyzed numerically. For spheres of small diameter (compared to the wavelength), the results coincide with the dipole approximation. We see that the proximity force approximation is not valid for spheres with diameters much larger than the gap, even though this approximation is well established for calculating forces. From the numerical results, a regime map is constructed based on two non--dimensional length scales for the validity of different approximations.
Necessary and sufficient conditions for large Nc equivalence between parent and daughter theories, for a wide class of orbifold projections of U(Nc) gauge theories, are just the natural requirements that the discrete symmetry used to define the projection not be spontaneously broken in the parent theory, and the discrete symmetry permuting equivalent gauge group factors not be spontaneously broken in the daughter theory. In this paper, we discuss the application of this result to Z_k projections of N=1 supersymmetric Yang-Mills theory in four dimensions, as well as various multi-flavor generalizations. Z_k projections with k > 2 yielding chiral gauge theories violate the symmetry realization conditions needed for large Nc equivalence, due to the spontaneous symmetry breaking of discrete chiral symmetry in the parent super-Yang-Mills theory. But for Z_2 projections, we show that previous assertions of large Nc inequivalence, in infinite volume, between the parent and daughter theories were based on incorrect mappings of vacuum energies, theta angles, or connected correlators between the two theories. With the correct identifications, there is no sign of any inconsistency. A subtl
The relation between equal-time and light-front wave functions is studied using models for which the four-dimensional solution of the Bethe-Salpeter wave function can be obtained. The popular prescription of defining the longitudinal momentum fraction using the instant-form free kinetic energy and third component of momentum is found to be incorrect except in the non-relativistic limit. The only presently known way to obtain light-front wave functions from rest-frame, instant-form wave functions is to boost the latter wave functions to the infinite momentum frame. Despite this fact, we prove a relation between certain integrals of the equal-time and light-front wave functions.
In this work, the plasmon coupling effect between two gold nanospheres on a gold slab is investigated. At plasmon resonance frequencies, electrons on the surface of the slab are absorbed into spheres and contribute to plasmon oscillation. This effect can help enhance the local electric field and optical coupling force between the two spheres.
Given a finite, flat morphism between embeddable noetherian schemes of pure dimension 1, we define the notion of direct and inverse image for generalized divisors and generalized line bundles. In the case when we deal with (possibly reducible, non-reduced) projective curves over a field and the codomain curve is smooth, we introduce the compactified Jacobians parametrizing torsion-free rank-1 sheaves and we study the Norm and inverse image maps between compactified Jacobians. Finally, we introduce and study the Prym stack defined as the kernel of the Norm map.
The two Journal Citation Reports of the Science Citation Index 2004 and the Social Science Citation Index 2004 were combined in order to analyze and map journals and specialties at the edges and in the overlap between the two databases. For journals which belong to the overlap (e.g., Scientometrics), the merger mainly enriches our insight into the structure which can be obtained from the two databases separately; but in the case of scientific journals which are more marginal in either database, the combination can provide a new perspective on the position and function of these journals (e.g., Environment and Planning B-Planning and Design). The combined database additionally enables us to map citation environments in terms of the various specialties comprehensively. Using the vector-space model, visualizations are provided for specialties that are parts of the overlap (information science, science & technology studies). On the basis of the resulting visualizations, "betweenness"--a measure from social network analysis--is suggested as an indicator for measuring the interdisciplinarity of journals.
We present a statistical approach for the discovery of relationships between mathematical entities that is based on linear regression and deep learning with fully connected artificial neural networks. The strategy is applied to computational knot data and empirical connections between combinatorial and hyperbolic knot invariants are revealed.
We derive heuristically approximate formulas for the negative $k$--moments $M_{-k}(x)$ of the gaps between consecutive primes$<x $ represented directly by $π(x)$ --- the number of primes up to $x$. In particular we propose an analytical formula for the sum of reciprocals of gaps between consecutive primes $<x : ~ M_{-1}(x)\sim \frac{π^2(x)}{x-2π(x)}\log\Big(\frac{x}{2π(x)}\Big) \sim x \log \log(x)/\log^2(x)$. We illustrate obtained results by the enormous computer data up to $x=4\times 10^{18}$.