1,720,992 research outputs found
Block layer decomposition schemes for training deep neural networks
No full textDeep feedforward neural networks’ (DFNNs) weight estimation relies on the solution of a very large nonconvex optimization problem that may have many local (no global) minimizers, saddle points and large plateaus. Furthermore, the time needed to find good solutions of the training problem heavily depends on both the number of samples and the number of weights (variables). In this work, we show how block coordinate descent (BCD) methods can be fruitful applied to DFNN weight optimization problem and embedded in online frameworks possibly avoiding bad stationary points. We first describe a batch BCD method able to effectively tackle difficulties due to the network’s depth; then we further extend the algorithm proposing an online BCD scheme able to scale with respect to both the number of variables and the number of samples. We perform extensive numerical results on standard datasets using various deep networks. We show that the application of BCD methods to the training problem of DFNNs improves over standard batch/online algorithms in the training phase guaranteeing good generalization performance as well
Case article — Production and distribution optimization of beach equipment for the marinero company
No full textWe present a mixed integer nonlinear mathematical programming model, covering a broad range of operations research (OR)-related topics. The case is designed to allow students to use knowledge acquired from OR and management science classes to model, analyze, and provide concrete solutions for the considered problem. Because of its high practicality, this exercise is an ideal tool to make the OR domain more accessible and to learn how to balance a problem’s complexity with the availability of algorithms for its solution. The case has been proposed as a competitive challenge and has been assigned both to students pursuing a bachelor’s degree in management engineering and students pursuing a master’s degree in industrial engineering. Students were grouped into working teams, and all teams competed against each other to get the best solution to win the challenge. Both the work-in-team and the challenge settings have been enjoyed by the students. During three lectures of 90 minutes, a brief review of OR-related tools and a detailed description and analysis of the case study have been provided to the students. Successive periodic debriefing meeting sessions have been scheduled to engage and monitor students during project development
Branching with hyperplanes in the criterion space: The frontier partitioner algorithm for biobjective integer programming
Full text linkWe present an algorithm for finding the complete Pareto frontier of biobjective integer programming problems. The method is based on the solution of a finite number of integer programs. The feasible sets of the integer programs are built from the original feasible set, by adding cuts that separate efficient solutions. Providing the existence of an oracle to solve suitably defined single objective integer subproblems, the algorithm can handle biobjective nonlinear integer problems, in particular biobjective convex quadratic integer optimization problems. Our numerical experience on a benchmark of biobjective integer linear programming instances shows the efficiency of the approach in comparison with existing state-of-the-art methods. Further experiments on biobjective integer quadratic programming instances are reported
Necessary and sufficient global optimality conditions for NLP reformulations of linear SDP problems
No full textIn this paper we consider the standard linear SDP problem, and its low rank
nonlinear programming reformulation, based on a Gramian representation of a positive semi-
definite matrix. For this nonconvex quadratic problem with quadratic equality constraints,
we give necessary and sufficient conditions of global optimality expressed in terms of the
Lagrangian function
An unconstrained minimization method for solving low rank SDP relaxations of the max cut problem
No full textIn this paper we consider low-rank semidefinite programming (LRSDP)
relaxations of combinatorial quadratic problems that are equivalent to the maxcut
problem. Using the Gramian representation of a positive semidefinite matrix, the
LRSDP problem can be formulated as the nonconvex nonlinear programming prob-
lem of minimizing a quadratic function with quadratic equality constraints. For the
solution of this problem we propose a continuously differentiable exact merit func-
tion that exploits the special structure of the constraints and we use this function to
define an efficient and globally convergent algorithm. Finally, we test our code on an
extended set of instances of the maxcut problem and we report comparisons with other
existing codes
Optimal siting and sizing of wayside energy storage systems in a D.C. railway line
Full text linkThe paper proposes an optimal siting and sizing methodology to design an energy storage system (ESS) for railway lines. The scope is to maximize the economic benefits. The problem of the optimal siting and sizing of an ESS is addressed and solved by a software developed by the authors using the particle swarm algorithm, whose objective function is based on the net present value (NPV). The railway line, using a standard working day timetable, has been simulated in order to estimate the power flow between the trains finding the siting and sizing of electrical substations and storage systems suitable for the railway network. Numerical simulations have been performed to test the methodology by assuming a new-generation of high-performance trains on a 3 kV direct current (d.c.) railway line. The solution found represents the best choice from an economic point of view and which allows less energy to be taken from the primary network
Doppler echocardiographic assessment of left ventricular diastolic fuction in acromegaly.
No full text
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
- …
