1,721,100 research outputs found
Il Problema del Caricamento di Pallet con Stratificazione
No full textIl Distributor’s Pallet Loading Problem (DPLP) è un problema di ottimizzazione combinatoria che si manifesta nei settori della logistica e del trasporto. Consiste nel disporre efficientemente un insieme di scatole tridimensionali, di varie dimensioni, forma e peso, su un numero finito di pallet, minimizzando il numero di pallet usati. La difficoltà aumenta quando, oltre ai vincoli "geometrici", vengono aggiunti vincoli del mondo reale, necessari per la sicurezza nella gestione e trasporto delle merci. I vincoli principali sono:
Stabilità: ogni layer di scatole deve fornire una base stabile per quelle impilate sopra, con sufficiente supporto di superficie.
Impilabilità: la differenza di altezza tra scatole di uno stesso layer deve restare entro limiti specifici per garantire la stabilità della struttura.
Compressione: ogni layer deve sopportare il peso cumulativo degli strati sovrastanti.
L'obiettivo del DPLP è impilare tutte le scatole sul minor numero possibile di pallet, rispettando sia i vincoli geometrici sia quelli reali. Questo problema è strongly NP-hard, poiché generalizza il classico problema del bin-packing. La soluzione del DPLP è cruciale per ridurre i costi di trasporto e l'impatto ambientale, poiché l'uso di meno pallet comporta un minor consumo di carburante e minori emissioni di CO2. Nonostante la rilevanza pratica di questo problema, poche ricerche si sono focalizzate sulla risoluzione con i vincoli reali sopra citati. Questa tesi mira a colmare tale lacuna esplorando diversi metodi esatti, euristici e supportati dal Machine Learning per sviluppare soluzioni efficienti al DPLP. Il primo approccio esatto proposto è un modello MILP (Mixed-Integer Linear Programming), che integra tutti i principali vincoli del problema. Tuttavia, all'aumentare della dimensione del problema, il modello MILP diventa computazionalmente oneroso, specialmente per grandi dataset.
Per affrontare queste difficoltà, viene introdotto un secondo metodo esatto basato sulla decomposizione combinatoria di Benders. Questo metodo scompone il problema in un problema principale e vari sottoproblemi, introducendo tagli per eliminare soluzioni non valide e accelerare il processo di ottimizzazione. Sebbene questo approccio riduca i tempi di calcolo in molti casi, il guadagno in efficienza non è garantito per tutte le istanze. Viene poi proposto un approccio euristico che utilizza algoritmi di bin-packing bidimensionali per creare layer di scatole. Questi layer vengono impilati sui pallet, garantendo il rispetto dei requisiti di stabilità e compressione. Sebbene i metodi euristici non forniscano sempre soluzioni ottimali, sono più rapidi, specialmente per istanze di grandi dimensioni. Per bilanciare la velocità delle euristiche con la precisione dei metodi esatti, viene presentato un approccio matheuristico. Questo metodo genera un set ridotto di layer fattibili tramite tecniche euristiche, che viene poi ottimizzato con un modello MILP per trovare la migliore disposizione. Per migliorare ulteriormente il metodo matheuristico, abbiamo sviluppato una variante avanzata che integra algoritmi di Machine Learning, come la Random Forest e la Support Vector Machine for Regression. Questi modelli sono addestrati per classificare e ordinare i layer in base al loro potenziale di far parte di una soluzione ottimale. Identificando i più promettenti già nelle prime fasi, si riduce lo spazio delle soluzioni, permettendo al modello MILP di concentrarsi su un sottoinsieme rilevante. Questo approccio arricchito dal Machine Learning migliora notevolmente l'efficienza computazionale nel risolvere il DPLP, soprattutto nei problemi su larga scala, senza compromettere la qualità della soluzione finale. Questa tesi contribuisce all'ottimizzazione delle operazioni logistiche e della supply chain, fornendo metodi che migliorano l'efficienza operativa, riducono i costi di trasporto e aumentano la sicurezza e l'affidabilità delle merci trasportate.The Distributor’s Pallet Loading Problem (DPLP) is a combinatorial optimization problem that arises in the logistics and transportation sectors. It involves the task of efficiently arranging a set of three-dimensional boxes, which vary in size, shape, and weight, onto a finite number of pallets while minimizing the number of pallets used. The problem becomes more challenging when we add to the above “geometrical” constraints various real-world constraints that are critical for the safe handling and transporting of goods. The constraints include:
Stability: Each layer of boxes must provide a stable base for the boxes stacked above by ensuring sufficient surface area support.
Stackability: The difference in height between boxes in a single layer must remain within certain limits to ensure that the overall structure is stackable.
Compression: Each layer must be able to support the cumulative weight of the layers stacked above it.
The goal of the DPLP is to ensure that all the boxes are packed onto the smallest possible number of pallets while satisfying the geometrical and these constraints. The problem is strongly NP-hard since it generalizes the classic bin-packing problem. Solving the DPLP is crucial for reducing transportation costs and environmental impacts, as fewer pallets lead to lower fuel consumption and CO2 emissions. Despite the practical importance of this problem, there has been little research focused on solving it with the real-world constraints mentioned above. This thesis aims to bridge that gap by exploring various exact, heuristic, and machine learning-enhanced methods to develop efficient solutions for the DPLP. The first exact approach we present is a MILP model, which directly integrates all major constraints of the problem. However, as the scale of the problem increases, the MILP model becomes computationally demanding, particularly for large datasets. To address these challenges, a second exact method based on combinatorial Benders' decomposition is introduced. This method breaks the problem into a master problem and multiple subproblems, introducing cuts to remove invalid solutions and accelerate the optimization process. While this method can reduce computational time in many cases, the efficiency gain is not guaranteed for all problem instances. A heuristic approach is presented that uses two-dimensional bin-packing algorithms to create layers of boxes. These layers are stacked on pallets while ensuring that stability and compression requirements are met. While heuristic methods do not always yield optimal solutions, they offer faster results, especially for larger instances. To balance the speed of heuristics and the precision of exact methods, a matheuristic approach is presented. This method generates a smaller set of feasible layers using heuristic techniques, which are then optimized using a MILP model to find the best stacking arrangement. To further enhance the matheuristic method, we developed a more advanced variant incorporating machine learning algorithms, such as Random Forest and Support Vector Machine for Regression. These machine learning models are trained to classify and rank layers based on their potential to form part of an optimal solution. By identifying the most promising layers early in the process, the solution space can be reduced, enabling the MILP model to focus on a smaller, more relevant subset of layers. This ML-augmented approach significantly improves the computational efficiency of solving the DPLP, particularly for large-scale problems, without compromising the quality of the final solution. The effectiveness of these methodologies is tested on randomly generated instances and real-world datasets. This thesis contributes to the optimization of logistics and supply chain operations by providing methods that not only improve operational efficiency but also reduce transportation costs and enhance the safety and reliability of palletized goods
Solving a Real-Life Distributor’s Pallet Loading Problem
Full text linkWe consider the distributor’s pallet loading problem where a set of different boxes are packed on the smallest number of pallets by satisfying a given set of constraints. In particular, we refer to a real-life environment where each pallet is loaded with a set of layers made of boxes, and both a stability constraint and a compression constraint must be respected. The stability requirement imposes the following: (a) to load at level k+1 a layer with total area (i.e., the sum of the bottom faces’ area of the boxes present in the layer) not exceeding α times the area of the layer of level k (where α≥1), and (b) to limit with a given threshold the difference between the highest and the lowest box of a layer. The compression constraint defines the maximum weight that each layer k can sustain; hence, the total weight of the layers loaded over k must not exceed that value. Some stability and compression constraints are considered in other works, but to our knowledge, none are defined as faced in a real-life problem. We present a matheuristic approach which works in two phases. In the first, a number of layers are defined using classical 2D bin packing algorithms, applied to a smart selection of boxes. In the second phase, the layers are packed on the minimum number of pallets by means of a specialized MILP model solved with Gurobi. Computational experiments on real-life instances are used to assess the effectiveness of the algorithm
Understanding community patterns in large attributed social networks
No full textThere is an inherent presence of communities in online social networks. These communities can be defined based on i) link structure or ii) the attributes of individuals. Attributes can indicate as interests in specific topics, like science-fiction books or romantic movies, or more in general their explicit affiliation to a group inside the network. In this paper, we analyze community structures as defined by how people are associated to third concepts like attributes. To understand the community patterns we analyze three large and one small social network datasets. Our analysis shows that, irrespective of the number of nodes for any particular interest in the network, at least 50% of the nodes are part of the same connected component in the graph induced by each interest. Another interesting result of our analysis is that the majority of sub-communities (50% or above) for any interest are separated by small hops (two to three) from each other
A Machine Learning Approach to Speed up the Solution of the Distributor’s Pallet Loading Problem
No full textPath-Based and Whole-Network Measures
No full textPath-based measures associate a value to every node in a network according to its direct
and indirect connections to other nodes. For example, given a node we can compute the
maximum distance to all other nodes: this measure is called node eccentricity. Whole-
network measures associate a value to an entire network, providing a summary of its
structure. For example, the diameter of a network is the maximum eccentricity of its
nodes, and represents a global measure of the efficiency of information dissemination in
that network. In this essay we cover the most popular path-based and whole-network
measures
Docking, 3D-QSAR studies and in silicoADME prediction on c-Src tyrosine kinase inhibitors
No full textDocking simulations and three-dimensional quantitative structure-activity relationship (3D-QSAR) analysis were performed on a wide set of c-Src inhibitors. The study was conducted using a structure-based alignment and by applying the GRID/GOLPE approach. The present 3D-QSAR investigation proved to be of good statistical value, displaying r(2), q(2) and cross-validation SDEP values of 0.94, 0.84 and 0.42, respectively. Moreover, such a model also proved to be capable of predicting the activities of an external test set of compounds. The availability of the 3D structure of the target made possible the interpretation of steric and electrostatic maps within the binding site environment and provided useful insight into the structural requirements for inhibitory activity against c-Src. Two regions whose occupation by hydrophobic portions of ligands would favourably affect the activity were clearly identified. Moreover, hydrogen bond interactions involving residues Met343, Asp406 and Ser347 emerged as playing a key role in determining the affinity of the active inhibitors toward c-Src. Furthermore, the inhibitors bearing a basic nitrogen provided enhanced potency through protonation and salt bridge formation with Asp350. A preliminary pharmacokinetic profile of the molecules under analysis was also drawn on the basis of Volsurf predictions
Investigating the types and effects of missing data in multilayer networks
No full textA common problem in social network analysis is the presence of missing data. This problem has been extensively investigated in single layer networks, that is, considering one network at a time. However, in multilayer networks, in which a holistic view of multiple networks is taken, the problem has not been specifically studied, and results for single layer networks are reused with no adaptation. In this work, we take an exhaustive and systematic approach to understand the effect of missing data in multilayer networks. Differently from the single layer networks, depending on layer interdependencies, the common network properties can increase or decrease with respect to the properties of the complete network. Another important aspect we observed through our experiments on real datasets is that multilayer network properties like layer correlation and relevance can be used to understand the impact of missing data compared to measuring traditional network measures
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
Variations on the Author
Full text link“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
- …
