Reports de recerca
http://hdl.handle.net/2117/3973
2016-08-29T07:20:08ZProcess mining meets abstract interpretation
http://hdl.handle.net/2117/88077
Process mining meets abstract interpretation
Carmona Vargas, Josep; Cortadella Fortuny, Jordi
The discovery of process models out of system traces is an interesting problem that has received significant attention in the last years. In this work, a theory for the derivation of a Petri net from a set of traces is presented. The method is based on the theory of abstract interpretation, which has been applied successfully in other areas. The principal application of the theory presented is Process Mining, an area that tries to incorporate the use of formal models both in the design and use of information systems.
2016-06-16T08:50:11ZCarmona Vargas, JosepCortadella Fortuny, JordiThe discovery of process models out of system traces is an interesting problem that has received significant attention in the last years. In this work, a theory for the derivation of a Petri net from a set of traces is presented. The method is based on the theory of abstract interpretation, which has been applied successfully in other areas. The principal application of the theory presented is Process Mining, an area that tries to incorporate the use of formal models both in the design and use of information systems.Comparing error minimized extreme learning machines and support vector sequential feed-forward neural networks
http://hdl.handle.net/2117/88076
Comparing error minimized extreme learning machines and support vector sequential feed-forward neural networks
Romero Merino, Enrique; Alquézar Mancho, René
Recently, error minimized extreme learning machines (EM-ELMs) have been proposed as a simple and efficient approach to build single-hidden-layer feed-forward networks (SLFNs) sequentially. They add random hidden nodes one by one (or group by group) and update the output weights incrementally to minimize the sum-of-squares error in the training set. Other very similar methods that also construct SLFNs sequentially had been reported earlier with the main difference that their hidden-layer weights are a subset of the data instead of being random. By analogy with the concept of support vectors original of support vector machines (SVMs), these approaches can be referred to as support vector sequential feed-forward neural networks (SV-SFNNs), and they are a particular case of the Sequential Approximation with Optimal Coefficients and Interacting Frequencies (SAOCIF) method. In this paper, it is firstly shown that EM-ELMs can also be cast as a particular case of SAOCIF. In particular, EM-ELMs can easily be extended to test some number of random candidates at each step and select the best of them, as SAOCIF does. Moreover, it is demonstrated that the cost of the calculation of the optimal output-layer weights in the originally proposed EM-ELMs can be improved if it is replaced by the one included in SAOCIF. Secondly, we present the results of an experimental study on 10 benchmark classification and 10 benchmark regression data sets, comparing EM-ELMs and SV-SFNNs, that was carried out under the same conditions for the two models. Although both models have the same (efficient) computational cost, a statistically significant improvement in generalization performance of SV-SFNNs vs. EM-ELMs was found in 12 out of the 20 benchmark problems.
2016-06-16T08:38:58ZRomero Merino, EnriqueAlquézar Mancho, RenéRecently, error minimized extreme learning machines (EM-ELMs) have been proposed as a simple and efficient approach to build single-hidden-layer feed-forward networks (SLFNs) sequentially. They add random hidden nodes one by one (or group by group) and update the output weights incrementally to minimize the sum-of-squares error in the training set. Other very similar methods that also construct SLFNs sequentially had been reported earlier with the main difference that their hidden-layer weights are a subset of the data instead of being random. By analogy with the concept of support vectors original of support vector machines (SVMs), these approaches can be referred to as support vector sequential feed-forward neural networks (SV-SFNNs), and they are a particular case of the Sequential Approximation with Optimal Coefficients and Interacting Frequencies (SAOCIF) method. In this paper, it is firstly shown that EM-ELMs can also be cast as a particular case of SAOCIF. In particular, EM-ELMs can easily be extended to test some number of random candidates at each step and select the best of them, as SAOCIF does. Moreover, it is demonstrated that the cost of the calculation of the optimal output-layer weights in the originally proposed EM-ELMs can be improved if it is replaced by the one included in SAOCIF. Secondly, we present the results of an experimental study on 10 benchmark classification and 10 benchmark regression data sets, comparing EM-ELMs and SV-SFNNs, that was carried out under the same conditions for the two models. Although both models have the same (efficient) computational cost, a statistically significant improvement in generalization performance of SV-SFNNs vs. EM-ELMs was found in 12 out of the 20 benchmark problems.The Virtual Magic Lantern: An interaction metaphor for enhanced medical data inspection
http://hdl.handle.net/2117/88074
The Virtual Magic Lantern: An interaction metaphor for enhanced medical data inspection
Monclús Lahoya, Eva; Díaz Iriberri, José; Navazo Álvaro, Isabel; Vázquez Alcocer, Pere Pau
In Volume Rendering, it is difficult to simultaneously visualize interior and exterior structures. Several approaches have been developed to solve this problem, such as cut-away or exploded views. Nevertheless, in most cases, those algorithms usually require either a preprocess of the data, or an accurate determination of the region of interest, previous to data inspection. In this paper we present the Virtual Magic Lantern (VML), an interaction tool tailored to facilitate volumetric data inspection. It behaves like a lantern whose virtual illumination cone provides the focal region which is visualized using a secondary transfer function or rendering style. This may be used for simple visual inspection, surgery planning, or injure diagnosis. The VML is a particularly friendly and intuitive interaction tool suitable for an immersive Virtual Reality setup with a large screen, where the user moves a Wanda device, like a lantern pointing to the model. We show that this inspection metaphor can be efficiently and easily adapted to a GPU ray casting volume visualization algorithm. We also present the Virtual Magic Window (VMW) metaphor as an efficient collateral implementation of the VML, that can be seen as a restricted case where the lantern illuminates following the viewing direction, through a virtual window created as the intersection of the virtual lantern (guided by the Wanda device) and the bounding box of the volume. This approach allows better frame rates because the shader requires almost no modification with respect to the classical volume ray casting. On the other hand, the contextual information is not perceived as well as with the previous method, especially on the boundary of the inspected region. Apart from its main use, our metaphor can also be used to aid in the tedious process of transfer function selection.
2016-06-16T08:27:55ZMonclús Lahoya, EvaDíaz Iriberri, JoséNavazo Álvaro, IsabelVázquez Alcocer, Pere PauIn Volume Rendering, it is difficult to simultaneously visualize interior and exterior structures. Several approaches have been developed to solve this problem, such as cut-away or exploded views. Nevertheless, in most cases, those algorithms usually require either a preprocess of the data, or an accurate determination of the region of interest, previous to data inspection. In this paper we present the Virtual Magic Lantern (VML), an interaction tool tailored to facilitate volumetric data inspection. It behaves like a lantern whose virtual illumination cone provides the focal region which is visualized using a secondary transfer function or rendering style. This may be used for simple visual inspection, surgery planning, or injure diagnosis. The VML is a particularly friendly and intuitive interaction tool suitable for an immersive Virtual Reality setup with a large screen, where the user moves a Wanda device, like a lantern pointing to the model. We show that this inspection metaphor can be efficiently and easily adapted to a GPU ray casting volume visualization algorithm. We also present the Virtual Magic Window (VMW) metaphor as an efficient collateral implementation of the VML, that can be seen as a restricted case where the lantern illuminates following the viewing direction, through a virtual window created as the intersection of the virtual lantern (guided by the Wanda device) and the bounding box of the volume. This approach allows better frame rates because the shader requires almost no modification with respect to the classical volume ray casting. On the other hand, the contextual information is not perceived as well as with the previous method, especially on the boundary of the inspected region. Apart from its main use, our metaphor can also be used to aid in the tedious process of transfer function selection.A multi-client architecture for hybrid terrain rendering on mobile devices
http://hdl.handle.net/2117/88073
A multi-client architecture for hybrid terrain rendering on mobile devices
Noguera, José María; Segura Sánchez, Rafael Jesús; Ogáyar, Carlos; Joan Arinyo, Robert
Mobile devices such as Personal Digital Assistants (PDAs) or smart phones are rapidly increasing their graphics and networking capabilities. However, real-time rendering of large terrains is still a challenging task to accomplish in such limited devices. In this paper we describe the principles involved in the design and development of a scalable client-server architecture for hybrid rendering of terrains over wireless networks on mobile devices. We have developed a hybrid adaptive streaming and rendering method based on a server-client approach. The rendering workload is distributed between a server and the clients and the terrain is partitioned into the close-range geometry and the background. The close-range geometry is downloaded from the database and rendered on the mobile client, and the background is portrayed as a view-dependent panoramic impostor and rendered by the server on demand then it is sent on request to the server for display. The system architecture is organized in three levels: the main server, the panorama server and the mobile device client. This architecture provides support for efficient delivery of geometry and impostors to mobile clients according to their capabilities. As a proof of concept, we have implemented a prototype and carried out exhaustive experiments considering different network scenarios and different number of connected clients. The analysis of the server workload and response times shows that our architecture achieves a great scalability and performance even when using low-end mobile devices.
2016-06-16T07:57:00ZNoguera, José MaríaSegura Sánchez, Rafael JesúsOgáyar, CarlosJoan Arinyo, RobertMobile devices such as Personal Digital Assistants (PDAs) or smart phones are rapidly increasing their graphics and networking capabilities. However, real-time rendering of large terrains is still a challenging task to accomplish in such limited devices. In this paper we describe the principles involved in the design and development of a scalable client-server architecture for hybrid rendering of terrains over wireless networks on mobile devices. We have developed a hybrid adaptive streaming and rendering method based on a server-client approach. The rendering workload is distributed between a server and the clients and the terrain is partitioned into the close-range geometry and the background. The close-range geometry is downloaded from the database and rendered on the mobile client, and the background is portrayed as a view-dependent panoramic impostor and rendered by the server on demand then it is sent on request to the server for display. The system architecture is organized in three levels: the main server, the panorama server and the mobile device client. This architecture provides support for efficient delivery of geometry and impostors to mobile clients according to their capabilities. As a proof of concept, we have implemented a prototype and carried out exhaustive experiments considering different network scenarios and different number of connected clients. The analysis of the server workload and response times shows that our architecture achieves a great scalability and performance even when using low-end mobile devices.Geometric constraint problems and solution instances
http://hdl.handle.net/2117/88071
Geometric constraint problems and solution instances
Hidalgo Garcia, Marta; Joan Arinyo, Robert; Soto Riera, Antoni
Geometric constraint solving is a growing field devoted to solve geometric problems defined by relationships, called constraints, established between the geometric elements. In this work we show that what characterizes a geometric constraint problem is the set of geometric elements on which the problem is defined. If the problem is wellconstrained, a given solution instance to the geometric constraint problem admits different representations defined by measuring geometric relationships in the solution instance.
2016-06-16T07:34:15ZHidalgo Garcia, MartaJoan Arinyo, RobertSoto Riera, AntoniGeometric constraint solving is a growing field devoted to solve geometric problems defined by relationships, called constraints, established between the geometric elements. In this work we show that what characterizes a geometric constraint problem is the set of geometric elements on which the problem is defined. If the problem is wellconstrained, a given solution instance to the geometric constraint problem admits different representations defined by measuring geometric relationships in the solution instance.Skeleton computation of an image using a geometric approach
http://hdl.handle.net/2117/88018
Skeleton computation of an image using a geometric approach
Martínez Bayona, Jonàs; Vigo Anglada, Marc; Pla García, Núria; Ayala Vallespí, M. Dolors
In this work we develop two algorithms to compute the skeleton of a binary 2D images. Both algorithms follow a geometric approach and work directly with the boundary of the image wich is an orthogonal polygon (OP). One of these algorithms processes the edges of the polygon while the other uses its vertices. Compared to a thinning method, the presented algorithms show a good performance.
2016-06-15T09:21:44ZMartínez Bayona, JonàsVigo Anglada, MarcPla García, NúriaAyala Vallespí, M. DolorsIn this work we develop two algorithms to compute the skeleton of a binary 2D images. Both algorithms follow a geometric approach and work directly with the boundary of the image wich is an orthogonal polygon (OP). One of these algorithms processes the edges of the polygon while the other uses its vertices. Compared to a thinning method, the presented algorithms show a good performance.Structural parameters
http://hdl.handle.net/2117/88016
Structural parameters
Ayala Vallespí, M. Dolors; Vergés, Eduard
The development of the bioengineering field has yield to the apparition of a new application of the Computer-Aided-Design, Bio-CAD, aiming to the use of computer models to represent biological samples and methods to simulate biological processes. Structural parameters allow to measure sample properties as osteoporosis degree of bones, suitability of biomaterials to be used as implants and transport and distribution of fluids into rocks for petrographic purposes among others. In this work a bibliographuc study is performed of structutal parameters mainly focussing on those related to the porosity and connectivity.
2016-06-15T08:56:08ZAyala Vallespí, M. DolorsVergés, EduardThe development of the bioengineering field has yield to the apparition of a new application of the Computer-Aided-Design, Bio-CAD, aiming to the use of computer models to represent biological samples and methods to simulate biological processes. Structural parameters allow to measure sample properties as osteoporosis degree of bones, suitability of biomaterials to be used as implants and transport and distribution of fluids into rocks for petrographic purposes among others. In this work a bibliographuc study is performed of structutal parameters mainly focussing on those related to the porosity and connectivity.A hybrid client-server based technique for navigation in large terrains using mobile devices
http://hdl.handle.net/2117/88015
A hybrid client-server based technique for navigation in large terrains using mobile devices
Noguera, José María; Segura Sánchez, Rafael Jesús; Ogáyar, Carlos; Joan Arinyo, Robert
We describe a hybrid client-server technique for remote adaptive streaming and rendering of large terrains in resource-limited mobile devices. The technique has been designed to achieve an interactive rendering performance on a mobile device connected to a low-bandwidth wireless network. The rendering workload is split between the client and the server. The terrain area close to the viewer is rendered in real-time by the client using a hierarchical multiresolution scheme. The terrain located far from the viewer is portrayed as view-dependent impostors, rendered by the server on demand and, then sent to the client. The hybrid technique provides tools to dynamically balance the rendering workload according to the resources available at the client side and to the saturation of the network and server. A prototype has been built and an exhaustive set of experiments covering several platforms, wireless networks and a wide range of viewer velocities has been conducted. Results show that the approach is feasible, effective and robust.
2016-06-15T08:47:35ZNoguera, José MaríaSegura Sánchez, Rafael JesúsOgáyar, CarlosJoan Arinyo, RobertWe describe a hybrid client-server technique for remote adaptive streaming and rendering of large terrains in resource-limited mobile devices. The technique has been designed to achieve an interactive rendering performance on a mobile device connected to a low-bandwidth wireless network. The rendering workload is split between the client and the server. The terrain area close to the viewer is rendered in real-time by the client using a hierarchical multiresolution scheme. The terrain located far from the viewer is portrayed as view-dependent impostors, rendered by the server on demand and, then sent to the client. The hybrid technique provides tools to dynamically balance the rendering workload according to the resources available at the client side and to the saturation of the network and server. A prototype has been built and an exhaustive set of experiments covering several platforms, wireless networks and a wide range of viewer velocities has been conducted. Results show that the approach is feasible, effective and robust.Process mining from a basis of state regions
http://hdl.handle.net/2117/88012
Process mining from a basis of state regions
Solé, Marc; Carmona Vargas, Josep
A central problem in the area of Process Mining is to obtain a formal model that represents selected behavior of a system. The theory of regions has been applied to address this problem, enabling the derivation of a Petri net whose language includes a set of traces. However, when dealing with real-life systems, the available tool support for performing such task is unsatisfactory, due to the complex algorithms that are required. In this paper, the theory of regions is revisited to devise a novel technique that explores the space of regions by combining the elements of a region basis. Due to its light space requirements, the approach can represent an important step for bridging the gap between the theory of regions and its industrial application. Experimental results improve in orders of magnitude state-of-the-art tools for the same task.
2016-06-15T07:59:02ZSolé, MarcCarmona Vargas, JosepA central problem in the area of Process Mining is to obtain a formal model that represents selected behavior of a system. The theory of regions has been applied to address this problem, enabling the derivation of a Petri net whose language includes a set of traces. However, when dealing with real-life systems, the available tool support for performing such task is unsatisfactory, due to the complex algorithms that are required. In this paper, the theory of regions is revisited to devise a novel technique that explores the space of regions by combining the elements of a region basis. Due to its light space requirements, the approach can represent an important step for bridging the gap between the theory of regions and its industrial application. Experimental results improve in orders of magnitude state-of-the-art tools for the same task.Multikey Quickselect
http://hdl.handle.net/2117/88009
Multikey Quickselect
Frias Moya, Leonor; Roura Ferret, Salvador
In this paper we introduce Multikey Quickselect: an efficient, in-place, and easy to implement algorithm for the selection problem for strings. We present several variants of our basic algorithm, which apply to two different flavors of the selection problem. Also, we analyze the cost of the main variants, measured as the expected number of character comparisons and elements swaps. Some of the enhancements presented in this paper apply to Multikey Quicksort as well.
2016-06-15T07:21:39ZFrias Moya, LeonorRoura Ferret, SalvadorIn this paper we introduce Multikey Quickselect: an efficient, in-place, and easy to implement algorithm for the selection problem for strings. We present several variants of our basic algorithm, which apply to two different flavors of the selection problem. Also, we analyze the cost of the main variants, measured as the expected number of character comparisons and elements swaps. Some of the enhancements presented in this paper apply to Multikey Quicksort as well.