Slideshow Image

Login Form

Research Areas

The main goal of LIDeCC consists in the design of new computational methodologies for solving real-world problems in Engineering and Bioinformatics.
Our developments are mostly based on Graph Theory, Evolutionary Computing and Machine Learning principles, combined with Numerical Methods and Parallel Processing techniques.
A brief description of our current basic and applied research lines is given below.

Basic Research Lines

Design of Multi-Objetive Evolutionary Algorithms

The inherent complexity associated to multi-objective optimization problems (MOPs) constitues a challenge for their resolution by means of deterministic algorithms, this being worst as the search space grows. In this context, evolutionary techniques have proven to be robust methods, effective for the resolution of MOPs, and popular as a consequence of their success. In particular, we are developing new knowledge on evolutionary techniques based on Pareto dominance. This will allow tackling, in a more effective way, lots of complex multi-objective optimization problems.

 

High Performance Algorithms for Optimization and Non-linear System Solving

In the field of Optimization, we are focused in developing parallel versions of the traditional sequential GRG and SQP algorithms, using a domain decomposition strategy. As regards the Non-linear System Solving problem, we are working on the design of a new methodology based on graph decomposition techniques and the master-worker paradigm.

 

Efficient Solving of equation systems

Solving systems of equations is a topic of interest in scientific computation because in many areas there is a great diversity of models defined by various systems of varying complexity. The state of the art on numerical optimization shows a breakthrough in the techniques employed to solve a growing range of application problems. There are algorithms with a very large theoretical support that make them strong and reliable at its implementation time. But in practice, it is known that those models tend to increase in complexity and size as its design becomes more realistic. Consequently, both the refinement of existing software processes and the design of new algorithms that support higher levels of complexity with minimal computational costs represent constant challenges. Currently, the main concern of researchers in this area of knowledge is to achieve robust strategies that solve rigorous models of large-scale problems in reduced execution times. The aim of this work is to develop a module that solves large and mixed systems, i.e. involving linear and nonlinear equations. As to execution times, this tool will make it possible to solve many complex problems of simulation and optimization more efficiently.

 

  

Geometry of Euclidean Hamiltonian Trajectories

Our research concerns the study of Euclidean Hamiltonian Cycles and Paths in the network built by a complete graph Kn with vertices on the n-th roots of the unity and with the Euclidean distances measured between nodes. The first aim is to single out and enumerate the reflective Euclidean Hamiltonian Cycles, whereas the second one is to find the optimum cycles and paths. A postulate of geometric optics allows us to identify the Shortest Euclidean Hamiltonian Cycles, and also the Longest Cycles only at odd cases. A methodology is developed. It singles out every longest path that solves each of the different (n/2) problems about the Longest Euclidean Hamiltonian Paths on the n−th root of the unity. This identification is done regardless planar rotations and orientation. A slightly modified technique is being applied in bilayered network architecture, as well as in a speci fic DNA folding problem.

 

Applied Research Lines

Process Engineering: Decision Support System for Instrumentation Design

The main goal of this research area is the development of decision support systems (DSS) in order to aid process engineers in the instrumentation design of industrial processes. This software integrates several algorithms and techniques for sensor network design developed by the LIDeCC staff. As a result, we expect to obtain a software package useful for the modern industry.

 

Process Engineering: Process Modelling and Simulation

We are working in the design of a middleware compatible with the CAPE-OPEN standard. This task constitutes a previous step for the development of plug-ins, which would allow the incorporation of tailor-made software simulation modules to industrial process simulation packages available in the international market.

 

Bioinformatics: Network Analysis

This research line's objective is to study machine learning and statistical techniques to infer gene regulatory networks and pathway networks from gene expression data. In particular, we are working in the inference of time delay associations among genes, in order to infer complex functional regulation behaviors, and multivariate analysis methods for the detection of common activity patterns between pathways.

 

Logistic Engineering: Optimization Methods for the Transit Network Problem

The bus-network scheduling problem (BNSP) is a NP-Complete problem, and the mathematical techniques for its resolution suffer difficulties in realistic sceneries. Several researchers have proposed heuristic alternatives for the BNSP. We are working on an hybrid approach that combines Greedy Randomized Adaptive Search Procedures (GRASP) and genetic algorithms, in order provide an effective computational tool for large-scale BNSP.

 

Bioinformatics: Prediction of ADMET Properties for Drug Design

Prediction of physicochemical properties is of major concern for pharmaceutical research. In this context, machine learning methods are of great importance due to their contribution to the development of a plethora of models. In particular, we are working in a novel framework for physicochemical property prediction, where different aspects of this problem - as molecular descriptors selection, applicability domain of the predictors and visual analytics of the data - are modeled.

 

  

Process Engineering: Metaheuristic Techniques Applied to the Optimal Design of Transport Networks for Natural-Gas Liquids

We are exploring the potential use of the liquids of natural gas as a petrochemical feedstock. The transport lines connecting oil fields in Santa Cruz are distributed through logistic tools. Modern metaheuristic techniques are proposed and a software package is being developed to achieve the optimal design.

 

Process Engineering: Exploitment of Natural Gas with high CO2 content

In a local context of great energy demand and growing prices of liquid and gaseous fuels, natural gas with high CO2 content represents a promising source to obtain synthesis gas, together with more valuable products coming from this source. Therefore, the transport networks and the possibility of expansion or design of process plants are being evaluated through economic analyses, digital georeferentiation, logistics and model building.

 

 

Computation of semantic similarity measures

Information systems use massive volumes of data that are commonly represented as matrices of considerable size and whose processing requires intensive computations. These matrices may be optimized accordingly to achieve an improvement in a target process. In particular, we aim at obtaining a more computationally efficient method in order to calculate semantic similarity between Web pages gathered in topic ontologies, such as Yahoo, Open Directory Project (ODP) and their derivatives.

 

Optimal Cycle Program of Traffic Lights

Pollution, congestion, security, parking, and many other problems derived from vehicular traffic are present every day in most cities around the world. Since changes in urban area infrastructure are usually not possible,a correct staging of traffic lights can help to reduce these problems by improving the flow of vehicles through the cities.