PyTorchKR Papers

최신 논문

총 1,463편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:1810.06545

A generalized GN-model closed-form formula

The GN-model of fiber non-linearity has had quite substantial success in modern optical telecommunications networks as a design and management tool. A version of it, capable of handling arbitrary WDM combs and link structures in closed form, was proposed in 2014. Here we upgrade that formula, to make it capable of handling frequency-dependent dispersion, frequency-dependent loss and frequency-dependent gain/loss due to Stimulated Raman Scattering (SRS) among channels. This way, more challenging and complex network scenarios, like the ones that are being deployed right now, can be dealt with in real-time, to the great advantage of management, control and optimization of such networks.

arXiv:1810.00771

A Preliminary Report on Probabilistic Attack Normal Form for Constellation Semantics

After Dung's founding work in Abstract Argumentation Frameworks there has been a growing interest in extending the Dung's semantics in order to describe more complex or real life situations. Several of these approaches take the direction of weighted or probabilistic extensions. One of the most prominent probabilistic approaches is that of constellation Probabilistic Abstract Argumentation Frameworks from Li~et~al. In this paper, we present a normal form for constellation probabilistic abstract argumentation frameworks. Furthermore, we present a transformation from general constellation probabilistic abstract argumentation frameworks to the presented normal form. In this way we illustrate that the simpler normal form has equal representation power with the general one.

arXiv:1810.00516

A New Form of Banking -- Concept and Mathematical Model of Venture Banking

This theoretical model contains concept, equations, and graphical results for venture banking. A system of 27 equations describes the behavior of the venture-bank and underwriter system allowing phase-space type graphs that show where profits and losses occur. These results confirm and expand those obtained from the original spreadsheet based model. An example investment in a castle at a loss is provided to clarify concept. This model requires that all investments are in enterprises that create new utility value. The assessed utility value created is the new money out of which the venture bank and underwriter are paid. The model presented chooses parameters that ensure that the venture-bank experiences losses before the underwriter does. Parameters are: DIN Premium, 0.05; Clawback lien fraction, 0.77; Clawback bonds and equity futures discount, 1.5 x (USA 12 month LIBOR); Range of clawback bonds sold, 0 to 100%; Range of equity futures sold 0 to 70%.

arXiv:1809.04520

Genetic algorithms with DNN-based trainable crossover as an example of partial specialization of general search

Universal induction relies on some general search procedure that is doomed to be inefficient. One possibility to achieve both generality and efficiency is to specialize this procedure w.r.t. any given narrow task. However, complete specialization that implies direct mapping from the task parameters to solutions (discriminative models) without search is not always possible. In this paper, partial specialization of general search is considered in the form of genetic algorithms (GAs) with a specialized crossover operator. We perform a feasibility study of this idea implementing such an operator in the form of a deep feedforward neural network. GAs with trainable crossover operators are compared with the result of complete specialization, which is also represented as a deep neural network. Experimental results show that specialized GAs can be more efficient than both general GAs and discriminative models.

arXiv:1808.07940

A Closed-Form Approximation of the Gaussian Noise Model in the Presence of Inter-Channel Stimulated Raman Scattering

An accurate, closed-form expression evaluating the nonlinear interference (NLI) power in coherent optical transmission systems in the presence of inter-channel stimulated Raman scattering (ISRS) is derived. The analytical result enables a rapid estimate of the signal-to-noise ratio (SNR) and avoids the need for integral evaluations and split-step simulations. The formula also provides new insight into the underlying parameter dependence of ISRS on the NLI. The proposed result is applicable for dispersion unmanaged, ultra-wideband transmission systems that use optical bandwidths of up to 15 THz. The accuracy of the closed-form expression is compared to numerical integration of the ISRS Gaussian Noise model and split-step simulations in a point-to-point transmission, as well as in a mesh optical network scenario.

arXiv:1807.10834

Estimating Diffeomorphic Mappings between Templates and Noisy Data: Variance Bounds on the Estimated Canonical Volume Form

Anatomy is undergoing a renaissance driven by availability of large digital data sets generated by light microscopy. A central computational task is to map individual data volumes to standardized templates. This is accomplished by regularized estimation of a diffeomorphic transformation between the coordinate systems of the individual data and the template, building the transformation incrementally by integrating a smooth flow field. The canonical volume form of this transformation is used to quantify local growth, atrophy, or cell density. While multiple implementations exist for this estimation, less attention has been paid to the variance of the estimated diffeomorphism for noisy data. Notably, there is an infinite dimensional un-observable space defined by those diffeomorphisms which leave the template invariant. These form the stabilizer subgroup of the diffeomorphic group acting on the template. The corresponding flat directions in the energy landscape are expected to lead to increased estimation variance. Here we show that a least-action principle used to generate geodesics in the space of diffeomorphisms connecting the subject brain to the template removes the stabilizer. This provides reduced-variance estimates of the volume form. Using simulations we demonstrate that the asymmetric large deformation diffeomorphic mapping methods (LDDMM), which explicitly incorporate the asymmetry between idealized template images and noisy empirical images, provide lower variance estimators than their symmetrized counterparts (cf. ANTs). We derive Cramer-Rao bounds for the variances in the limit of small deformations. Analytical results are shown for the Jacobian in terms of perturbations of the vector fields and divergence of the vector field.

arXiv:1807.10619

Power Minimizer Symbol-Level Precoding: A Closed-Form Sub-Optimal Solution

In this letter, we study the optimal solution of the multiuser symbol-level precoding (SLP) for minimization of the total transmit power under given signal-to-interference-plus-noise ratio (SINR) constraints. Adopting the distance preserving constructive interference regions (DPCIR), we first derive a simplified reformulation of the problem. Then, we analyze the structure of the optimal solution using the Karush-Kuhn-Tucker (KKT) optimality conditions, thereby we obtain the necessary and sufficient condition under which the power minimizer SLP is equivalent to the conventional zero-forcing beamforming (ZFBF). This further leads us to a closed-form sub-optimal SLP solution (CF-SLP) for the original problem. Simulation results show that CF-SLP provides significant gains over ZFBF, while performing quite close to the optimal SLP in scenarios with rather small number of users. The results further indicate that the CF-SLP method has a reduction of order $10^3$ in computational time compared to the optimal solution.

arXiv:1805.01631

Is Information in the Brain Represented in Continuous or Discrete Form?

The question of continuous-versus-discrete information representation in the brain is a fundamental yet unresolved question. Historically, most analyses assume a continuous representation without considering the discrete alternative. Our work explores the plausibility of both, answering the question from a communications systems engineering perspective. Using Shannon's communications theory, we posit that information in the brain is represented in discrete form. We address this hypothesis using 2 approaches. First, we identify the fundamental communication requirements of the brain. Second, we estimate the symbol error probability and channel capacity for a continuous information representation. Our work concludes that information cannot be communicated and represented reliably in the brain using a continuous representation - it has to be in a discrete form. This is a major demarcation from conventional and current wisdom. We apply this discrete result to the 4 major neural coding hypotheses, and illustrate the use of discrete ISI neural coding in analyzing electrophysiology experimental data. We further posit and illustrate a plausible direct link between Weber's Law and discrete neural coding. We end by outlining a number of key research questions on discrete neural coding.

arXiv:1804.08904

Closed-form approximations in derivatives pricing: The Kristensen-Mele approach

Kristensen and Mele (2011) developed a new approach to obtain closed-form approximations to continuous-time derivatives pricing models. The approach uses a power series expansion of the pricing bias between an intractable model and some known auxiliary model. Since the resulting approximation formula has closed-form it is straightforward to obtain approximations of greeks. In this thesis I will introduce Kristensen and Mele's methods and apply it to a variety of stochastic volatility models of European style options as well as a model for commodity futures. The focus of this thesis is the effect of different model choices and different model parameter values on the numerical stability of Kristensen and Mele's approximation.

arXiv:1803.11096

Adaptive Parameters Adjustment for Group Reweighted Zero-Attracting LMS

Group zero-attracting LMS and its reweighted form have been proposed for addressing system identification problems with structural group sparsity in the parameters to estimate. Both algorithms however suffer from a trade-off between sparsity degree and estimation bias and, in addition, between convergence speed and steady-state performance like most adaptive filtering algorithms. It is therefore necessary to properly set their step size and regularization parameter. Based on a model of their transient behavior, we introduce a variable-parameter variant of both algorithms to address this issue. By minimizing their mean-square deviation at each time instant, we obtain closed-form expressions of the optimal step size and regularization parameter. Simulation results illustrate the effectiveness of the proposed algorithms.

arXiv:1803.07574

Water Residence Time Estimation by 1D Deconvolution in the Form of a l2-Regularized Inverse Problem With Smoothness, Positivity and Causality Constraints

The Water Residence Time distribution is the equivalent of the impulse response of a linear system allowing the propagation of water through a medium, e.g. the propagation of rain water from the top of the mountain towards the aquifers. We consider the output aquifer levels as the convolution between the input rain levels and the Water Residence Time, starting with an initial aquifer base level. The estimation of Water Residence Time is important for a better understanding of hydro-bio-geochemical processes and mixing properties of wetlands used as filters in ecological applications, as well as protecting fresh water sources for wells from pollutants. Common methods of estimating the Water Residence Time focus on cross-correlation, parameter fitting and non-parametric deconvolution methods. Here we propose a 1D full-deconvolution, regularized, non-parametric inverse problem algorithm that enforces smoothness and uses constraints of causality and positivity to estimate the Water Residence Time curve. Compared to Bayesian non-parametric deconvolution approaches, it has a fast runtime per test case; compared to the popular and fast cross-correlation method, it produces a more precise Water Residence Time curve even in the case of noisy measurements. The algorithm needs only one regularization parameter to balance between smoothness of the Water Residence Time and accuracy of the reconstruction. We propose an approach on how to automatically find a suitable value of the regularization parameter from the input data only. Tests on real data illustrate the potential of this method to analyze hydrological datasets.

arXiv:1802.09425

Plant species accumulation curves are determined by evenness and spatial aggregation in drylands worldwide

Species accumulation curves (SAC), i.e. the relationship between species richness and the number of sampling units in a given community, can be used to describe diversity patterns while accounting for the well-known scale-dependence of species richness. Despite their value, the functional form and the parameters of SAC, as well as their determinants, have barely been investigated in plant communities, particularly in drylands. We characterized the SAC of perennial plant communities from 233 dryland ecosystems from six continents by comparing the fit of major functions (power-law, logarithmic and Michaelis-Menten). We tested the theoretical prediction that the effects of aridity and soil pH on SAC are mediated by vegetation attributes such as evenness, cover, and spatial aggregation. We found that the logarithmic relationship was the most common functional form, followed by Michaelis-Menten and power-law. Functional form was mainly determined by evenness while the SAC parameters (intercept and slope) were largely determined by spatial aggregation. In addition, aridity decreased small scale richness (intercept of SAC) but did not affect accumulation rate (slope of the SAC). Our results highlight the role that attributes such as spatial aggregation and evenness play as main mediators of the SAC of vegetation in drylands, the Earth's largest biome.

arXiv:1802.07741

Extended Reduced-Form Framework for Non-Life Insurance

In this paper we propose a general framework for modeling an insurance liability cash flow in continuous time, by generalizing the reduced-form framework for credit risk and life insurance. In particular, we assume a nontrivial dependence structure between the reference filtration and the insurance internal filtration. We apply these results for pricing and hedging non-life insurance liabilities in hybrid financial and insurance markets, while taking into account the role of inflation under the benchmarked risk-minimization approach. This framework offers at the same time a general and flexible structure, and an explicit and treatable pricing-hedging formula.

arXiv:1801.06028

A closed-form formula for pricing bonds between coupon payments

We derive a closed-form formula for computing bond prices between coupon payments. Our results cover both the `Treasury' and the `Street' pricing methods used by sovereign and corporate issuers. We apply our formulas to two UK gilts, the 8% Treasury Gilt 2015, and the 0.5% Treasury Gilt 2022, and show that we can obtain the dirty price of these bonds at any date with a minimum of calculations, and without intensive computational resources.

arXiv:1709.06517

Numerical analysis for a unified 2 factor model of structural and reduced form types for corporate bonds with fixed discrete coupon

Conditions of Stability for explicit finite difference scheme and some results of numerical analysis for a unified 2 factor model of structural and reduced form types for corporate bonds with fixed discrete coupon are provided. It seems to be difficult to get solution formula for PDE model which generalizes Agliardi's structural model [1] for discrete coupon bonds into a unified 2 factor model of structural and reduced form types and we study a numerical analysis for it by explicit finite difference scheme. These equations are parabolic equations with 3 variables and they include mixed derivatives, so the explicit finite difference scheme is not stable in general. We find conditions for the explicit finite difference scheme to be stable, in the case that it is stable, numerically compute the price of the bond and analyze its credit spread and duration.

arXiv:1707.04475

Reduced-form framework under model uncertainty

In this paper we introduce a sublinear conditional expectation with respect to a family of possibly nondominated probability measures on a progressively enlarged filtration. In this way, we extend the classic reduced-form setting for credit and insurance markets to the case under model uncertainty, when we consider a family of priors possibly mutually singular to each other. Furthermore, we study the superhedging approach in continuous time for payment streams under model uncertainty, and establish several equivalent versions of dynamic robust superhedging duality. These results close the gap between robust framework for financial market, which is recently studied in an intensive way, and the one for credit and insurance markets, which is limited in the present literature only to some very specific cases.

arXiv:1704.04553

Trilobite "pelotons": Possible hydrodynamic drag effects between leading and following trilobites in trilobite queues

Energy saving mechanisms in nature allow following organisms to expend less energy than leaders. Queues, or ordered rows of individuals, may form when organisms exploit the available energy saving mechanism while travelling at near-maximal sustainable metabolic capacities; compact clusters form when group members travel well below maximal sustainable metabolic capacities. The group size range, given here as the ratio of the difference between the size of the largest and smallest group members, and the size of the largest member (as a per cent), has been hypothesized to correspond proportionately to the energy saving quantity because weaker, smaller, individuals sustain speeds of stronger, larger, individuals by exploiting the energy saving mechanism (as a per cent). During migration, small individuals outside this range may perish, or form sub-groups, or simply not participate in migratory behavior. We approximate drag forces for leading and following individuals in queues of the late Devonian (approx. 370 Ma) trilobite Trimerocephalus chopini. Applying data from literature of R. herculea, a living crustacean, we approximate the hypothetical walking speed and maximal speeds for T. chopini. Findings reasonably support the hypothesis: among the population of fossilized queues of T. chopini reported by Kin and Blazejowski (2013), trilobite size range was 75% while the size range within queues, was 63%; this corresponds reasonably with drag reductions in following positions that permit approx. 61.5% energy saving for trilobites following others in optimal low-drag positions. We model collective trilobite behavior associated with hydrodynamic drafting.

arXiv:1703.10481

DNA Tweezers Based on Semantics of DNA Strand Graph

Because of the limitations of classical silicon based computational technology, several alternatives to traditional method in form of unconventional computing have been proposed. In this paper we will focus on DNA computing which is showing the possibility of excellence for its massive parallelism, potential for information storage, speed and energy efficiency. In this paper we will describe how syllogistic reasoning by DNA tweezers can be presented by the semantics of process calculus and DNA strand graph. Syllogism is an essential ingredient for commonsense reasoning of an individual. This paper enlightens the procedure to deduce a precise conclusion from a set of propositions by using formal language theory in form of process calculus and the expressive power of DNA strand graph.

arXiv:1612.05660

Gr\"obner Bases of Neural Ideals

The brain processes information about the environment via neural codes. The neural ideal was introduced recently as an algebraic object that can be used to better understand the combinatorial structure of neural codes. Every neural ideal has a particular generating set, called the canonical form, that directly encodes a minimal description of the receptive field structure intrinsic to the neural code. On the other hand, for a given monomial order, any polynomial ideal is also generated by its unique (reduced) Gr\"obner basis with respect to that monomial order. How are these two types of generating sets -- canonical forms and Gr\"obner bases -- related? Our main result states that if the canonical form of a neural ideal is a Gr\"obner basis, then it is the universal Gr\"obner basis (that is, the union of all reduced Gr\"obner bases). Furthermore, we prove that this situation -- when the canonical form is a Gr\"obner basis -- occurs precisely when the universal Gr\"obner basis contains only pseudo-monomials (certain generalizations of monomials). Our results motivate two questions: (1)~When is the canonical form a Gr\"obner basis? (2)~When the universal Gr\"obner basis of a neural ideal is {\em not} a canonical form, what can the non-pseudo-monomial elements in the basis tell us about the receptive fields of the code? We give partial answers to both questions. Along the way, we develop a representation of pseudo-monomials as hypercubes in a Boolean lattice.

arXiv:1610.04754

Three-dimensional contact of TITH cartilage layers, a closed-form solution

Inhomogeneity and anisotropy play a crucial role in attributing articular cartilage its properties. The frictionless contact model constructed here consists in two thin biphasic transversely isotropic transversely homogeneous (TITH) cartilage layers firmly attached onto rigid substrates and shaped as elliptic paraboloids of different radii. Using asymptotic techniques, a solution to the deformation problem of such material has been recently obtained extending previous ones referred to homogeneous materials. The layer itself is thin in comparison with the size of the contact area and the observed time is shorter than the hydrogel characteristic time. The emerging three-dimensional contact problem is solved in closed-form and numerical benchmarks for constant and oscillating loads are given. The results are shown in terms of contact pressure and approach of the bones. The latter is derived to be directly proportional to the contact area. Existing experimental data are reinterpreted in view of the current model formulation. Comparisons are made with existing solutions for homogeneous biphasic materials in order to underline the functional importance of inhomogeneity in spreading the contact pressure distribution across the contact area. Particular attention is paid to the applicability of the retrieved formulas for interpreting measurements of in vivo experiments. Future directions are also prospected.