Showing 1–27 of 27 results for author: Cho, K

Recursive Self-Composite Approach Towards Structural Understanding of Boolean Network

Authors: Jongrae Kim, Woojeong Lee, Kwang-Hyun Cho

Abstract: Boolean networks have been widely used in many areas of science and engineering to represent various dynamical behaviour. In systems biology, they became useful tools to study the dynamical characteristics of large-scale biomolecular networks and there have been a number of studies to develop efficient ways of finding steady states or cycles of Boolean network models. On the other hand, there has… ▽ More Boolean networks have been widely used in many areas of science and engineering to represent various dynamical behaviour. In systems biology, they became useful tools to study the dynamical characteristics of large-scale biomolecular networks and there have been a number of studies to develop efficient ways of finding steady states or cycles of Boolean network models. On the other hand, there has been little attention to analyzing the dynamic properties of the network structure itself. Here, we present a systematic way to study such properties by introducing a recursive self-composite of the logic update rules. Of note, we found that all Boolean update rules actually have repeated logic structures underneath. This repeated nature of Boolean networks reveals interesting algebraic properties embedded in the networks. We found that each converged logic leads to the same states, called kernel states. As a result, the longest-length period of states cycle turns out to be equal to the number of converged logics in the logic cycle. Based on this, we propose a lea** and filling algorithm to avoid any possible large string explosions during the self-composition procedures. Finally, we demonstrate how the proposed approach can be used to reveal interesting hidden properties using Boolean network examples of a simple network with a long feedback structure, a T-cell receptor network and a cancer network. △ Less

Submitted 28 December, 2023; originally announced December 2023.

Comments: 9 pages, 3 figures

Betty: An Automatic Differentiation Library for Multilevel Optimization

Authors: Sang Keun Choe, Willie Neiswanger, Pengtao Xie, Eric Xing

Abstract: Gradient-based multilevel optimization (MLO) has gained attention as a framework for studying numerous problems, ranging from hyperparameter optimization and meta-learning to neural architecture search and reinforcement learning. However, gradients in MLO, which are obtained by composing best-response Jacobians via the chain rule, are notoriously difficult to implement and memory/compute intensive… ▽ More Gradient-based multilevel optimization (MLO) has gained attention as a framework for studying numerous problems, ranging from hyperparameter optimization and meta-learning to neural architecture search and reinforcement learning. However, gradients in MLO, which are obtained by composing best-response Jacobians via the chain rule, are notoriously difficult to implement and memory/compute intensive. We take an initial step towards closing this gap by introducing Betty, a software library for large-scale MLO. At its core, we devise a novel dataflow graph for MLO, which allows us to (1) develop efficient automatic differentiation for MLO that reduces the computational complexity from O(d^3) to O(d^2), (2) incorporate systems support such as mixed-precision and data-parallel training for scalability, and (3) facilitate implementation of MLO programs of arbitrary complexity while allowing a modular interface for diverse algorithmic and systems design choices. We empirically demonstrate that Betty can be used to implement an array of MLO programs, while also observing up to 11% increase in test accuracy, 14% decrease in GPU memory usage, and 20% decrease in training wall time over existing implementations on multiple benchmarks. We also showcase that Betty enables scaling MLO to models with hundreds of millions of parameters. We open-source the code at https://github.com/leopard-ai/betty. △ Less

Submitted 15 March, 2023; v1 submitted 5 July, 2022; originally announced July 2022.

Optimal Control of Several Motion Models

Authors: Tan H. Cao, Nilson Chapagain, Kangmin Cho, **woo Choi, Sinae Hong, Abhishek Kafle, Haejoon Lee, Hansol Lim, Biniam Markos, Jiung Seo, Phung Ngoc Thi, Nguyen Nang Thieu

Abstract: This paper is devoted to the study of the dynamic optimization of several controlled crowd motion models in the general planar settings, which is an application of a class of optimal control problems involving a general nonconvex swee** process with perturbations. A set of necessary optimality conditions for such optimal control problems involving the crowd motion models with multiple agents and… ▽ More This paper is devoted to the study of the dynamic optimization of several controlled crowd motion models in the general planar settings, which is an application of a class of optimal control problems involving a general nonconvex swee** process with perturbations. A set of necessary optimality conditions for such optimal control problems involving the crowd motion models with multiple agents and obstacles is obtained and analyzed. Several effective algorithms based on such necessary optimality conditions are proposed and various nontrivial illustrative examples together with their simulations are also presented. △ Less

Submitted 30 April, 2022; originally announced May 2022.

MSC Class: 49J52; 49J53; 49K24; 49M25; 90C30

Simple quadrature rules for a nonparametric nonconforming quadrilateral element

Authors: Kanghun Cho, Dongwoo Sheen

Abstract: We introduce simple quadrature rules for the family of nonparametric nonconforming quadrilateral element with four degrees of freedom. Our quadrature rules are motivated by the work of Meng {\it et al.} \cite{meng2018new}. First, we introduce a family of MVP (Mean Value Property)-preserving four DOFs nonconforming elements on the intermediate reference domain introduced by Meng {\it et al.}. Then… ▽ More We introduce simple quadrature rules for the family of nonparametric nonconforming quadrilateral element with four degrees of freedom. Our quadrature rules are motivated by the work of Meng {\it et al.} \cite{meng2018new}. First, we introduce a family of MVP (Mean Value Property)-preserving four DOFs nonconforming elements on the intermediate reference domain introduced by Meng {\it et al.}. Then we design two--points and three--points quadrature rules on the intermediate reference domain. Under the assumption on equal quadrature weights, the deviation from the quadrilateral center of the Gauss points for the two points and three points rules assumes the same quadratic polynomials with constant terms modified. Thus, the two--points rule and three--points rule are constructed at one stroke. The quadrature rules are asymptotically optimal as the mesh size is sufficiently small. Several numerical experiments are carried out, which show efficiency and convergence properties of the new quadrature rules. △ Less

Submitted 25 January, 2022; originally announced January 2022.

MSC Class: 65N30

Algebraic Multiscale Method for one--dimensional elliptic problems

Authors: Kanghun Cho, Roktaek Lim, Dongwoo Sheen

Abstract: In this paper we propose an idea of constructing a macro--scale matrix system given a micro--scale matrix linear system. Then the macro--scale system is solved at cheaper computing costs. The method uses the idea of the generalized multiscale finite element method based. Some numerical results are presented. In this paper we propose an idea of constructing a macro--scale matrix system given a micro--scale matrix linear system. Then the macro--scale system is solved at cheaper computing costs. The method uses the idea of the generalized multiscale finite element method based. Some numerical results are presented. △ Less

Submitted 25 January, 2022; originally announced January 2022.

MSC Class: 65N30

Algebraic Multiscale Method for two--dimensional elliptic problems

Authors: Kanghun Cho, Imbunm Kim, Raehyun Kim, Dongwoo Sheen

Abstract: We introduce an algebraic multiscale method for two--dimensional problems. The method uses the generalized multiscale finite element method based on the quadrilateral nonconforming finite element spaces. Differently from the one--dimensional algebraic multiscale method, we apply the dimension reduction techniques to construct multiscale basis functions. Also moment functions are considered to impo… ▽ More We introduce an algebraic multiscale method for two--dimensional problems. The method uses the generalized multiscale finite element method based on the quadrilateral nonconforming finite element spaces. Differently from the one--dimensional algebraic multiscale method, we apply the dimension reduction techniques to construct multiscale basis functions. Also moment functions are considered to impose continuity between local basis functions. Some representative numerical results are presented. △ Less

Submitted 25 January, 2022; originally announced January 2022.

MSC Class: 65N30

Federated Adaptive Causal Estimation (FACE) of Target Treatment Effects

Authors: Larry Han, Jue Hou, Kelly Cho, Rui Duan, Tianxi Cai

Abstract: Federated learning of causal estimands may greatly improve estimation efficiency by leveraging data from multiple study sites, but robustness to heterogeneity and model misspecifications is vital for ensuring validity. We develop a Federated Adaptive Causal Estimation (FACE) framework to incorporate heterogeneous data from multiple sites to provide treatment effect estimation and inference for a f… ▽ More Federated learning of causal estimands may greatly improve estimation efficiency by leveraging data from multiple study sites, but robustness to heterogeneity and model misspecifications is vital for ensuring validity. We develop a Federated Adaptive Causal Estimation (FACE) framework to incorporate heterogeneous data from multiple sites to provide treatment effect estimation and inference for a flexibly specified target population of interest. FACE accounts for site-level heterogeneity in the distribution of covariates through density ratio weighting. To safely incorporate source sites and avoid negative transfer, we introduce an adaptive weighting procedure via a penalized regression, which achieves both consistency and optimal efficiency. Our strategy is communication-efficient and privacy-preserving, allowing participating sites to share summary statistics only once with other sites. We conduct both theoretical and numerical evaluations of FACE and apply it to conduct a comparative effectiveness study of BNT162b2 (Pfizer) and mRNA-1273 (Moderna) vaccines on COVID-19 outcomes in U.S. veterans using electronic health records from five VA regional sites. We show that compared to traditional methods, FACE meaningfully increases the precision of treatment effect estimates, with reductions in standard errors ranging from $26\%$ to $67\%$. △ Less

Submitted 5 October, 2023; v1 submitted 16 December, 2021; originally announced December 2021.

Comments: 59 pages

Density of Periodic Points for Lattès maps over Finite Fields

Authors: Zoë Bell, Jasmine Camero, Karina Cho, Trevor Hyde, Chieh-Mi Lu, Rebecca Miller, Bianca Thompson, Eric Zhu

Abstract: Let $L_d$ be the Lattès map associated to the multiplication-by-$d$ endomorphism of an elliptic curve $E$ defined over a finite field $\mathbb{F}_q$. We determine the density $δ(L_d,q)$ of periodic points for $L_d$ in $\mathbb{P}^1(\mathbb{F}_q)$. We show that the periodic point densities $δ(L_d,q^n)$ converge as $n \rightarrow \infty$ along certain arithmetic progressions, and compute simple expl… ▽ More Let $L_d$ be the Lattès map associated to the multiplication-by-$d$ endomorphism of an elliptic curve $E$ defined over a finite field $\mathbb{F}_q$. We determine the density $δ(L_d,q)$ of periodic points for $L_d$ in $\mathbb{P}^1(\mathbb{F}_q)$. We show that the periodic point densities $δ(L_d,q^n)$ converge as $n \rightarrow \infty$ along certain arithmetic progressions, and compute simple explicit formulas for $δ(L_\ell,q)$ when $\ell$ is a prime and $E$ belongs to a special family of supersingular elliptic curves. △ Less

Submitted 26 February, 2021; originally announced March 2021.

Comments: 13 pages

Dichotomy between Deterministic and Probabilistic Models in Countably Additive Effectus Theory

Authors: Kenta Cho, Bas Westerbaan, John van de Wetering

Abstract: Effectus theory is a relatively new approach to categorical logic that can be seen as an abstract form of generalized probabilistic theories (GPTs). While the scalars of a GPT are always the real unit interval [0,1], in an effectus they can form any effect monoid. Hence, there are quite exotic effectuses resulting from more pathological effect monoids. In this paper we introduce sigma-effectus… ▽ More Effectus theory is a relatively new approach to categorical logic that can be seen as an abstract form of generalized probabilistic theories (GPTs). While the scalars of a GPT are always the real unit interval [0,1], in an effectus they can form any effect monoid. Hence, there are quite exotic effectuses resulting from more pathological effect monoids. In this paper we introduce sigma-effectuses, where certain countable sums of morphisms are defined. We study in particular sigma-effectuses where unnormalized states can be normalized. We show that a non-trivial sigma-effectus with normalization has as scalars either the two-element effect monoid 0,1 or the real unit interval [0,1]. When states and/or predicates separate the morphisms we find that in the 0,1 case the category must embed into the category of sets and partial functions (and hence the category of Boolean algebras), showing that it implements a deterministic model, while in the [0,1] case we find it embeds into the category of Banach order-unit spaces and of Banach pre-base-norm spaces (satisfying additional properties), recovering the structure present in GPTs. Hence, from abstract categorical and operational considerations we find a dichotomy between deterministic and convex probabilistic models of physical theories. △ Less

Submitted 5 September, 2021; v1 submitted 23 March, 2020; originally announced March 2020.

Comments: In Proceedings QPL 2020, arXiv:2109.01534

Journal ref: EPTCS 340, 2021, pp. 91-113

Effectuses in Categorical Quantum Foundations

Authors: Kenta Cho

Abstract: This thesis develops the theory of effectuses as a categorical axiomatic approach to quantum theory. It provides a comprehensive introduction to effectus theory and reveals its connections with various other topics and approaches. This thesis develops the theory of effectuses as a categorical axiomatic approach to quantum theory. It provides a comprehensive introduction to effectus theory and reveals its connections with various other topics and approaches. △ Less

Submitted 27 October, 2019; originally announced October 2019.

Comments: PhD thesis, Radboud University (2019)

Quandle Cocycle Quivers

Authors: Karina Cho, Sam Nelson

Abstract: We incorporate quandle cocycle information into the quandle coloring quivers we defined in arXiv:1807.10465 to define weighted directed graph-valued invariants of oriented links we call \textit{quandle cocycle quivers}. This construction turns the quandle cocycle invariant into a small category, yielding a categorification of the quandle cocycleinvariant. From these graphs we define several new li… ▽ More We incorporate quandle cocycle information into the quandle coloring quivers we defined in arXiv:1807.10465 to define weighted directed graph-valued invariants of oriented links we call \textit{quandle cocycle quivers}. This construction turns the quandle cocycle invariant into a small category, yielding a categorification of the quandle cocycleinvariant. From these graphs we define several new link invariants including a 2-variable polynomial which specializes to the usual quandle cocycle invariant. Examples and computations are provided. △ Less

Submitted 19 April, 2019; originally announced April 2019.

Comments: 9 pages

MSC Class: 57M25; 57M27

Duplicable von Neumann Algebras

Authors: Kenta Cho, Abraham A. Westerbaan

Abstract: Recently, we have shown that von Neumann algebras form a model for Selinger and Valiron's quantum lambda calculus. In this paper, we explain our choice of interpretation of the duplicability operator "!" by studying those von Neumann algebras that might have served as the interpretation of duplicable types, namely those that carry a (commutative) monoid structure with respect to the spatial tensor… ▽ More Recently, we have shown that von Neumann algebras form a model for Selinger and Valiron's quantum lambda calculus. In this paper, we explain our choice of interpretation of the duplicability operator "!" by studying those von Neumann algebras that might have served as the interpretation of duplicable types, namely those that carry a (commutative) monoid structure with respect to the spatial tensor product. We show that every such monoid is the (possibly infinite) direct product of the complex numbers, and that our interpretation of the ! operator of the quantum lambda calculus is exactly the free (commutative) monoid. △ Less

Submitted 6 March, 2019; originally announced March 2019.

Comments: arXiv admin note: text overlap with arXiv:1804.02203

Quandle Coloring Quivers

Authors: Karina Cho, Sam Nelson

Abstract: We consider a quiver structure on the set of quandle colorings of an oriented knot or link diagram. This structure contains a wealth of knot and link invariants and provides a categorification of the quandle counting invariant in the most literal sense, i.e., giving the set of quandle colorings the structure of a small category which is unchanged by Reidemeister moves. We derive some new enhanceme… ▽ More We consider a quiver structure on the set of quandle colorings of an oriented knot or link diagram. This structure contains a wealth of knot and link invariants and provides a categorification of the quandle counting invariant in the most literal sense, i.e., giving the set of quandle colorings the structure of a small category which is unchanged by Reidemeister moves. We derive some new enhancements of the counting invariant from this quiver structure and show that the enhancements are proper with explicit examples. △ Less

Submitted 5 October, 2018; v1 submitted 27 July, 2018; originally announced July 2018.

Comments: 11 pages. Version 2 includes typo corrections suggested by referee

MSC Class: 57M27; 57M25

Generalizations of Triangle Inequalities to Spherical and Hyperbolic Geometry

Authors: Karina Cho, Jacob Naranjo

Abstract: Certain triangle inequalities involving the circumradius, inradius, and side lengths of a triangle are generalized to spherical and hyperbolic geometry. Examples include strengthenings of Euler's inequality, $R\geq2r$. An extension of Euler's inequality to a simplex in $n$-dimensional space is also generalized to spherical geometry. Certain triangle inequalities involving the circumradius, inradius, and side lengths of a triangle are generalized to spherical and hyperbolic geometry. Examples include strengthenings of Euler's inequality, $R\geq2r$. An extension of Euler's inequality to a simplex in $n$-dimensional space is also generalized to spherical geometry. △ Less

Submitted 26 May, 2018; originally announced May 2018.

Comments: Based on work done at 2017 Oregon State University Mathematics REU. 11 pages, 2 figures

MSC Class: 51M10

Indistinguishable binomial decision tree of 3-SAT: Proof of class P is a proper subset of class NP

Authors: Keum-Bae Cho

Abstract: This paper solves a long standing open problem of whether NP-complete problems could be solved in polynomial time on a deterministic Turing machine by showing that the indistinguishable binomial decision tree can be formed in a 3-SAT instance. This paper describes how to construct the decision tree and explains why 3-SAT has no polynomial-time algorithm when the decision tree is formed in the 3-SA… ▽ More This paper solves a long standing open problem of whether NP-complete problems could be solved in polynomial time on a deterministic Turing machine by showing that the indistinguishable binomial decision tree can be formed in a 3-SAT instance. This paper describes how to construct the decision tree and explains why 3-SAT has no polynomial-time algorithm when the decision tree is formed in the 3-SAT instance. The indistinguishable binomial decision tree consists of polynomial numbers of nodes containing an indistinguishable variable pair but generates exponentially many paths connecting the clauses to be used for sequences of resolution steps. The number of paths starting from the root node and arriving at a child node forms a binomial coefficient. In addition, each path has an indistinguishable property from one another. Due to the exponential number of paths and their indistinguishability, if an indistinguishable binomial decision tree is constructed in which there exist one or more paths generating an empty clause, the number of calculation steps needed to extract the empty clause is not polynomially bounded. This result leads to the conclusion that class P is a proper subset of class NP. △ Less

Submitted 30 January, 2018; originally announced January 2018.

Comments: 17 pages, 9 figures

Existence of mixed type solutions in the Chern-Simons gauge theory of rank two in $\mathbb{R}^2$

Authors: Kwangseok Choe, Namkwon Kim, Youngae Lee, Chang-Shou Lin

Abstract: We consider the Chern-Simons gauge theory of rank $2$ such as $SU(3)$, $SO(5)$, and $G_2$ Chern-Simons model in $\mathbb{R}^2$. There may exist three types of solutions in these theories, that is, topological, nontopological, and mixed type solutions. Among others, mixed type solutions can only exist in non-Abelian Chern-Simons models. We show the existence of mixed type solutions with an arbitrar… ▽ More We consider the Chern-Simons gauge theory of rank $2$ such as $SU(3)$, $SO(5)$, and $G_2$ Chern-Simons model in $\mathbb{R}^2$. There may exist three types of solutions in these theories, that is, topological, nontopological, and mixed type solutions. Among others, mixed type solutions can only exist in non-Abelian Chern-Simons models. We show the existence of mixed type solutions with an arbitrary configuration of vortex points which has been a long-standing open problem. To show it, as the first step, we need to find when a priori bound would fail. For the purpose, we shall find partially blowing up mixed type solutions by using different scalings for different components. Due to the different scalings, we should control the mass contribution from infinity which is one of the important parts in this paper. △ Less

Submitted 16 June, 2017; originally announced June 2017.

Comments: 22 pages

Von Neumann Algebras form a Model for the Quantum Lambda Calculus

Authors: Kenta Cho, Abraham Westerbaan

Abstract: We present a model of Selinger and Valiron's quantum lambda calculus based on von Neumann algebras, and show that the model is adequate with respect to the operational semantics. We present a model of Selinger and Valiron's quantum lambda calculus based on von Neumann algebras, and show that the model is adequate with respect to the operational semantics. △ Less

Submitted 7 March, 2016; originally announced March 2016.

Total and Partial Computation in Categorical Quantum Foundations

Authors: Kenta Cho

Abstract: This paper uncovers the fundamental relationship between total and partial computation in the form of an equivalence of certain categories. This equivalence involves on the one hand effectuses, which are categories for total computation, introduced by Jacobs for the study of quantum/effect logic. On the other hand, it involves what we call FinPACs with effects; they are finitely partially additive… ▽ More This paper uncovers the fundamental relationship between total and partial computation in the form of an equivalence of certain categories. This equivalence involves on the one hand effectuses, which are categories for total computation, introduced by Jacobs for the study of quantum/effect logic. On the other hand, it involves what we call FinPACs with effects; they are finitely partially additive categories equipped with effect algebra structures, serving as categories for partial computation. It turns out that the Kleisli category of the lift monad (-)+1 on an effectus is always a FinPAC with effects, and this construction gives rise to the equivalence. Additionally, state-and-effect triangles over FinPACs with effects are presented. △ Less

Submitted 4 November, 2015; originally announced November 2015.

Comments: In Proceedings QPL 2015, arXiv:1511.01181

Journal ref: EPTCS 195, 2015, pp. 116-135

Semantics for a Quantum Programming Language by Operator Algebras

Authors: Kenta Cho

Abstract: This paper presents a novel semantics for a quantum programming language by operator algebras, which are known to give a formulation for quantum theory that is alternative to the one by Hilbert spaces. We show that the opposite category of the category of W*-algebras and normal completely positive subunital maps is an elementary quantum flow chart category in the sense of Selinger. As a consequenc… ▽ More This paper presents a novel semantics for a quantum programming language by operator algebras, which are known to give a formulation for quantum theory that is alternative to the one by Hilbert spaces. We show that the opposite category of the category of W*-algebras and normal completely positive subunital maps is an elementary quantum flow chart category in the sense of Selinger. As a consequence, it gives a denotational semantics for Selinger's first-order functional quantum programming language QPL. The use of operator algebras allows us to accommodate infinite structures and to handle classical and quantum computations in a unified way. △ Less

Submitted 29 December, 2014; originally announced December 2014.

Comments: In Proceedings QPL 2014, arXiv:1412.8102

Journal ref: EPTCS 172, 2014, pp. 165-190

The number system hidden inside the Boolean satisfiability problem

Authors: Keum-Bae Cho

Abstract: This paper gives a novel approach to analyze SAT problem more deeply. First, I define new elements of Boolean formula such as dominant variable, decision chain, and chain coupler. Through the analysis of the SAT problem using the elements, I prove that we can construct a k-SAT (k>2) instance where the coefficients of cutting planes take exponentially large values in the input size. This exponentia… ▽ More This paper gives a novel approach to analyze SAT problem more deeply. First, I define new elements of Boolean formula such as dominant variable, decision chain, and chain coupler. Through the analysis of the SAT problem using the elements, I prove that we can construct a k-SAT (k>2) instance where the coefficients of cutting planes take exponentially large values in the input size. This exponential property is caused by the number system formed from the calculation of coefficients. In addition, I show that 2-SAT does not form the number system and Horn-SAT partially forms the number system according to the feasible value of the dominant variable. Whether or not the coefficients of cutting planes in cutting plane proof are polynomially bounded was open problem. Many researchers believed that cutting plane proofs with large coefficients are highly non-intuitive20. However, we can construct a k-SAT (k>2) instance in which cutting planes take exponentially large coefficients by the number system. In addition, this exponential property is so strong that it gives definite answers for several questions: why Horn-SAT has the intermediate property between 2-SAT and 3-SAT; why random-SAT is so easy; and why k-SAT (k>2) cannot be solved with the linear programming technique. As we know, 2-SAT is NL-complete, Horn-SAT is P-complete, and k-SAT (k>2) is NP-complete. In terms of computational complexity, this paper gives a clear mathematical property by which SAT problems in three different classes are distinguished. △ Less

Submitted 24 January, 2018; v1 submitted 17 June, 2014; originally announced June 2014.

Comments: 15 pages, 9 figures

Identifying and attacking the saddle point problem in high-dimensional non-convex optimization

Authors: Yann Dauphin, Razvan Pascanu, Caglar Gulcehre, Kyunghyun Cho, Surya Ganguli, Yoshua Bengio

Abstract: A central challenge to many fields of science and engineering involves minimizing non-convex error functions over continuous, high dimensional spaces. Gradient descent or quasi-Newton methods are almost ubiquitously used to perform such minimizations, and it is often thought that a main source of difficulty for these local methods to find the global minimum is the proliferation of local minima wit… ▽ More A central challenge to many fields of science and engineering involves minimizing non-convex error functions over continuous, high dimensional spaces. Gradient descent or quasi-Newton methods are almost ubiquitously used to perform such minimizations, and it is often thought that a main source of difficulty for these local methods to find the global minimum is the proliferation of local minima with much higher error than the global minimum. Here we argue, based on results from statistical physics, random matrix theory, neural network theory, and empirical evidence, that a deeper and more profound difficulty originates from the proliferation of saddle points, not local minima, especially in high dimensional problems of practical interest. Such saddle points are surrounded by high error plateaus that can dramatically slow down learning, and give the illusory impression of the existence of a local minimum. Motivated by these arguments, we propose a new approach to second-order optimization, the saddle-free Newton method, that can rapidly escape high dimensional saddle points, unlike gradient descent and quasi-Newton methods. We apply this algorithm to deep or recurrent neural network training, and provide numerical evidence for its superior optimization performance. △ Less

Submitted 10 June, 2014; originally announced June 2014.

Comments: The theoretical review and analysis in this article draw heavily from arXiv:1405.4604 [cs.LG]

Asymptotic analysis of oscillatory integrals via the Newton polyhedra of the phase and the amplitude

Authors: Koji Cho, Joe Kamimoto, Toshihiro Nose

Abstract: The asymptotic behavior at infinity of oscillatory integrals is in detail investigated by using the Newton polyhedra of the phase and the amplitude. We are especially interested in the case that the amplitude has a zero at a critical point of the phase. The properties of poles of local zeta functions, which are closely related to the behavior of oscillatory integrals, are also studied under the as… ▽ More The asymptotic behavior at infinity of oscillatory integrals is in detail investigated by using the Newton polyhedra of the phase and the amplitude. We are especially interested in the case that the amplitude has a zero at a critical point of the phase. The properties of poles of local zeta functions, which are closely related to the behavior of oscillatory integrals, are also studied under the associated situation. △ Less

Submitted 4 March, 2012; originally announced March 2012.

Comments: 36 pages

MSC Class: 58K55 (Primary) 14B05; 14M25 (Secondary)

Journal ref: J. Math. Soc. Japan, 65, No. 2 (2013) 521-562

On a multi-point Schwarz-Pick lemma

Authors: Kyung Hyun Cho, Seong-A Kim, Toshiyuki Sugawa

Abstract: We consider the multi-point Schwarz-Pick lemma and its associate functions due to Beardon-Minda and Baribeau-Rivard-Wegert. Basic properties of the associate functions are summarized. Then we observe that special cases of the multi-point Schwarz-Pick lemma give Schur's continued fraction algorithm and several inequalities for bounded analytic functions on the unit disk. We consider the multi-point Schwarz-Pick lemma and its associate functions due to Beardon-Minda and Baribeau-Rivard-Wegert. Basic properties of the associate functions are summarized. Then we observe that special cases of the multi-point Schwarz-Pick lemma give Schur's continued fraction algorithm and several inequalities for bounded analytic functions on the unit disk. △ Less

Submitted 1 February, 2011; originally announced February 2011.

Comments: 14 pages

MSC Class: Primary 30C80; Secondary 30F45; 53A35

Baker-Akhiezer Modules on the Intersections of Shifted Theta Divisors

Authors: Koji Cho, Andrey Mironov, Atsushi Nakayashiki

Abstract: The restriction, on the spectral variables, of the Baker-Akhiezer (BA) module of a g-dimensional principally polarized abelian variety with the non-singular theta divisor to an intersection of shifted theta divisors is studied. It is shown that the restriction to a k-dimensional variety becomes a free module over the ring of differential operators in $k$ variables. The remaining g-k derivations de… ▽ More The restriction, on the spectral variables, of the Baker-Akhiezer (BA) module of a g-dimensional principally polarized abelian variety with the non-singular theta divisor to an intersection of shifted theta divisors is studied. It is shown that the restriction to a k-dimensional variety becomes a free module over the ring of differential operators in $k$ variables. The remaining g-k derivations define evolution equations for generators of the BA-module. As a corollary new examples of commutative ring of partial differential operators with matrix coefficients and their non-trivial evolution equations are obtained. △ Less

Submitted 23 April, 2010; originally announced April 2010.

Comments: 14 pages

MSC Class: 14K25; 14H70; 35Q51

Triple Hilbert transforms along polynomial surfaces

Authors: Yong Kum Cho, Sunggeum Hong, Joonil Kim, Chan Woo Yang

Abstract: This paper has been withdrawn since it contains some discrepancy with othe authers's recent result. We will not post this until this discrepancy is resolved. This paper has been withdrawn since it contains some discrepancy with othe authers's recent result. We will not post this until this discrepancy is resolved. △ Less

Submitted 24 July, 2007; v1 submitted 20 July, 2007; originally announced July 2007.

Comments: This paper has been withdrawn

MSC Class: 42B20

Uniqueness and examples of compact toric Sasaki-Einstein metrics

Authors: Koji Cho, Akito Futaki, Hajime Ono

Abstract: In [11] it was proved that, given a compact toric Sasaki manifold of positive basic first Chern class and trivial first Chern class of the contact bundle, one can find a deformed Sasaki structure on which a Sasaki-Einstein metric exists. In the present paper we first prove the uniqueness of such Einstein metrics on compact toric Sasaki manifolds modulo the action of the identity component of the… ▽ More In [11] it was proved that, given a compact toric Sasaki manifold of positive basic first Chern class and trivial first Chern class of the contact bundle, one can find a deformed Sasaki structure on which a Sasaki-Einstein metric exists. In the present paper we first prove the uniqueness of such Einstein metrics on compact toric Sasaki manifolds modulo the action of the identity component of the automorphism group for the transverse holomorphic structure, and secondly remark that the result of [11] implies the existence of compatible Einstein metrics on all compact Sasaki manifolds obtained from the toric diagrams with any height, or equivalently on all compact toric Sasaki manifolds whose cones have flat canonical bundle. We further show that there exists an infinite family of inequivalent toric Sasaki-Einstein metrics on $S^5 \sharp k(S^2 \times S^3)$ for each positive integer $k$. △ Less

Submitted 10 January, 2007; v1 submitted 3 January, 2007; originally announced January 2007.

Comments: Statements of the results are modified

MSC Class: 53C55; 53C21; 55N91

Journal ref: Commun.Math.Phys.277:439-458,2008

Differential Structure of Abelian Functions

Authors: K. Cho, A. Nakayashiki

Abstract: The space of abelian functions of a principally polarized abelian variety J is studied as a module over the ring D of global holomorphic differential operators on J. We construct a D-free resolution in case the theta divisor is non-singular. As an application, in the case of dimension 2 and 3, we construct a new linear basis of the space of abelian functions in terms of logarithmic derivatives o… ▽ More The space of abelian functions of a principally polarized abelian variety J is studied as a module over the ring D of global holomorphic differential operators on J. We construct a D-free resolution in case the theta divisor is non-singular. As an application, in the case of dimension 2 and 3, we construct a new linear basis of the space of abelian functions in terms of logarithmic derivatives of the higher dimensional sigma function. △ Less

Submitted 12 April, 2006; originally announced April 2006.

Comments: 25 pages

MSC Class: 14K25 (primary) 14H40; 14H70 (secondary)