Recent studies in using deep learning (DL) to solve routing problems focus on construction heuristics, whose solutions are still far from optimality. Improvement heuristics have great potential narrow this gap by iteratively refining a solution. However, classic improvement all guided handcrafted rules that may limit their performance. In article, we propose reinforcement framework learn the for problems. We design self-attention-based architecture as policy network guide selection of next...

Existing deep reinforcement learning (DRL)-based methods for solving the capacitated vehicle routing problem (CVRP) intrinsically cope with a homogeneous fleet, in which fleet is assumed as repetitions of single vehicle. Hence, their key to construct solution solely lies selection next node (customer) visit excluding However, vehicles real-world scenarios are likely be heterogeneous different characteristics that affect capacity (or travel speed), rendering existing DRL less effective. In...

To facilitate queries over semi-structured data, various structural summaries have been proposed. Structural are derived directly from the data and serve as indices for evaluating path expressions on or XML data. We introduce D(k) index, an adaptive summary general graph structured documents. Building previous work, 1-index A(k) D(k)-index is also based concept of bisimilarity. However, a generalization A(k)-index, index possesses ability to adjust its structure according current query load....

In this paper, we propose a metaheuristic to solve the pickup and delivery problem with time windows. Our approach is tabu-embedded simulated annealing algorithm which restarts search procedure from current best solution after several non-improving iterations. The computational experiments on six newly-generated different data sets marked our as first large multiple-vehicle PDPTW instances various distribution properties.

In this paper, we propose a metaheuristic to solve the pickup and delivery problem with time windows. Our approach is tabu-embedded simulated annealing algorithm which restarts search procedure from current best solution after several non-improving iterations. The computational experiments on six newly-generated different data sets marked our as first large multiple-vehicle PDPTW instances various distribution properties.

We present and evaluate the capacity of a deep neural network to learn robust features from EEG automatically detect seizures. This is challenging problem because seizure manifestations on are extremely variable both inter- intra-patient. By simultaneously capturing spectral, temporal spatial information our recurrent convolutional learns general spatially invariant representation seizure. The proposed approach exceeds significantly previous results obtained cross-patient classifiers in...

The container relocation problem, where containers that are stored in bays retrieved a fixed sequence, is crucial port operation. Existing approaches using branch and bound algorithms only able to optimally solve small cases practical time frame. In this paper, we investigate iterative deepening A* (rather than bound) new lower measures heuristics, show approach much larger instances of the problem frame suitable for application. We also examine more difficult variant has been largely...

Recently, there is an emerging trend to apply deep reinforcement learning solve the vehicle routing problem (VRP), where a learnt policy governs selection of next node for visiting. However, existing methods could not handle well pairing and precedence relationships in pickup delivery (PDP), which representative variant VRP. To address this challenging issue, we leverage novel neural network integrated with heterogeneous attention mechanism empower automatically select nodes. In particular,...

While deep learning in 3D domain has achieved revolutionary performance many tasks, the robustness of these models not been sufficiently studied or explored. Regarding adversarial samples, most existing works focus on manipulation local points, which may fail to invoke global geometry properties, like under linear projection that preserves Euclidean distance, i.e., isometry. In this work, we show state-of-the-art are extremely vulnerable isometry transformations. Armed with Thompson...

Abstract In this work, we examine port crane scheduling with spatial and separation constraints. Although common to most operations, these constraints have not been previously studied. We assume that cranes cannot cross, there is a minimum distance between jobs be done simultaneously. The objective find crane‐to‐job matching which maximizes throughput under provide dynamic programming algorithms, probabilistic tabu search, squeaky wheel optimization heuristic for solution. Experiments show...