一种基于小字典不对等语料的跨语言词嵌入方法

双语词嵌入通常采用从源语言空间到目标语言空间映射,通过源语言映射嵌入到目标语言空间的最小距离线性变换实现跨语言词嵌入。然而大型的平行语料难以获得,词嵌入的准确率难以提高。针对语料数量不对等、双语语料稀缺情况下的跨语言词嵌入问题,该文提出一种基于小字典不对等语料的跨语言词嵌入方法,首先对单语词向量进行归一化,对小字典词对正交最优线性变换求得梯度下降初始值,然后通过对大型源语言(英语)语料进行聚类,借助小字典找到与每一聚类簇相对应的源语言词,取聚类得到的每一簇词向量均值和源语言与目标语言对应的词向量均值,建立新的双语词向量对应关系,将新建立的双语词向量扩展到小字典中,使得小字典得以泛化和扩展。最后,利用泛化扩展后的字典对跨语言词嵌入映射模型进行梯度下降求得最优值。在英语—意大利语、德语和芬兰语上进行了实验验证,实验结果证明该文方法可以在跨语言词嵌入中减少梯度下降迭代次数,减少训练时间,同时在跨语言词嵌入上表现出较好的正确率。     

步长自适应的测量矩阵迭代优化方法

在压缩感知中，降低传感矩阵的列相干性可以提高重构精度。因为稀疏字典一般是固定的，所以目前主要通过优化测量矩阵来间接降低传感矩阵列相干性。提出一种改进的测量矩阵优化算法，使用梯度下降法更新测量矩阵并结合Barzilai-Borwen方法以及Armijo准则，使步长能够在迭代中自适应调整并保证算法收敛性。仿真实验表明，所提出的方法具有更快的收敛速度并且能够得到更优的测量矩阵。     

Optimal Pricing Algorithm in the Electricity Market with Battery and Accumulator and Demand–Supply Balancing

This paper considers the day‐ahead market with batteries and accumulators to level power generation. First, we model consumers with batteries, generators with batteries, and accumulators, all of whom strive to maximize their own profit. An optimal pricing algorithm based on dual decomposition and the steepest descent method is derived and shown to be stable. Finally, the effectiveness of the algorithm is demonstrated by means of numerical simulations which indicate that it achieves demand–supply balance.     

Large neighborhood search for the bike request scheduling problem

This paper introduces an efficient algorithm for the bike request scheduling problem (BRSP). The BRSP is built around the concept of request, defined as the pickup or dropoff of a number of identical items (bikes) at a specific station, within a certain time window, and with a certain priority. The aim of the BRSP is to sequence requests on (and hence determine the routes of) a set of vehicles, in such a way that the sum of the priorities of the executed requests is maximized, all time windows are respected, and the capacity of the vehicles is not exceeded. The generation of the set of requests is explicitly not a part of the problem definition of the BRSP. The primary application of the BRSP, from which it derives its name, is to determine the routes of a set of repositioning vehicles in a bike sharing system, although other applications exist. The algorithm introduced in this paper is based on a set of related greedy randomized adaptive search procedure followed by variable neighborhood descent (GRASP + VND) operators embedded in a large neighborhood search (LNS) framework. Since this paper presents the first heuristic for the BRSP, a computational comparison to existing approaches is not possible. We therefore compare the solutions found by our LNS heuristic to those found by an exact solver (Gurobi). These experiments confirm that the proposed algorithm scales to realistic dimensions and is able to find near‐optimal solutions in seconds.     

Fix-and-relax approaches for controlled tabular adjustment

Controlled tabular adjustment (CTA) is a relatively new protection technique for tabular data protection. CTA formulates a mixed integer linear programming problem, which is challenging for tables of moderate size. Even finding a feasible initial solution may be a challenging task for large instances. On the other hand, end users of tabular data protection techniques give priority to fast executions and are thus satisfied in practice with suboptimal solutions. This work has two goals. First, the fix-and-relax (FR) strategy is applied to obtain good feasible initial solutions to large CTA instances. FR is based on partitioning the set of binary variables into clusters to selectively explore a smaller branch-and-cut tree. Secondly, the FR solution is used as a warm start for a block coordinate descent (BCD) heuristic (approach named FR+BCD); BCD was confirmed to be a good option for large CTA instances in an earlier paper by the second and third co-authors (Comput Oper Res 2011;38:1826–35 [23]). We report extensive computational results on a set of real-world and synthetic CTA instances. FR is shown to be competitive compared to CPLEX branch-and-cut in terms of quickly finding either a feasible solution or a good upper bound. FR+BCD improved the quality of FR solutions for approximately 25% and 50% of the synthetic and real-world instances, respectively. FR or FR+BCD provided similar or better solutions in less CPU time than CPLEX for 73% of the difficult real-world instances.     

A novel approach to fuzzy wavelet neural network modeling and optimization

In this paper, an efficient approach of combining Takagi–Sugeno–Kang fuzzy system with wavelet based neural network is presented. The model replaces the constant or a linear function of inputs in conclusion part of traditional TSK fuzzy model with wavelet neural network (WNN), thus each rule uses fuzzy set to separate the input space into subspaces spanned by different wavelet functions. For finding the optimal values for parameters of our proposed fuzzy wavelet neural network (proposed-FWNN), a hybrid learning algorithm integrating an improved particle swarm optimization (PSO) and gradient descent algorithm is employed. The two-layer inline-PSO process is proposed in this paper, whose adjustment scheme is more fitting the consequent pattern learning based gradient descent optimization and will locate a good region in the search space. Simulation examples are given to test the efficiency of proposed-FWNN model for identification of the dynamic plants. It is seen that our modeling and optimization approach results in a better performance.     

A multiobjective GRASP–VND algorithm to solve the waste collection problem

In this paper, the waste collection problem (WCP) of a city in the south of Spain is addressed as a multiobjective routing problem that considers three objectives. From the company's perspective, the minimization of the travel cost is desired as well as that of the total number of vehicles. Additionally, from the employee's point of view, a set of balanced routes is also sought. Four variants of a multiobjective hybrid algorithm are proposed. Specifically, a GRASP (greedy randomized adaptive search procedure) with a VND (variable neighborhood descent) is combined. The best GRASP–VND algorithm found is applied in order to solve the real‐world WCP of a city in the south of Spain.     

Efficient and High-quality Recommendations via Momentum-incorporated Parallel Stochastic Gradient Descent-Based Learning

A recommender system (RS) relying on latent factor analysis usually adopts stochastic gradient descent (SGD) as its learning algorithm. However, owing to its serial mechanism, an SGD algorithm suffers from low efficiency and scalability when handling large-scale industrial problems. Aiming at addressing this issue, this study proposes a momentum-incorporated parallel stochastic gradient descent (MPSGD) algorithm, whose main idea is two-fold: a) implementing parallelization via a novel data-splitting strategy, and b) accelerating convergence rate by integrating momentum effects into its training process. With it, an MPSGD-based latent factor (MLF) model is achieved, which is capable of performing efficient and high-quality recommendations. Experimental results on four high-dimensional and sparse matrices generated by industrial RS indicate that owing to an MPSGD algorithm, an MLF model outperforms the existing state-of-the-art ones in both computational efficiency and scalability.      

Dynamic modeling and optimal control of goethite process based on the rate-controlling step

The iron removal process is an important technology in zinc hydrometallurgy. Due to its complicated reaction mechanism, it is difficult to highly control the performance of process solely relying on manual experience. Therefore, this paper focuses on the dynamic modeling for the goethite process and its optimal control method. In different reactors, different operating conditions will influence the rate-controlling step of process reactions, as well as the kinetic model. We determine the rate-controlling step of reactions in each reactor according to the reaction conditions, and subsequently the dynamic model has been developed based on the rate-controlling step. Then an optimal control method for the goethite process has been proposed, which satisfies the technical requirements with minimal process consumption. The proposed optimal control includes pre-setting of descent gradient of outlet ferrous ion concentration and optimal control of oxygen and zinc oxide. The simulation results demonstrate that the proposed dynamic model exhibits greater performance comparing with the process model without considering rate-controlling step, and the proposed control strategy has a higher satisfactory than the nonlinear model predictive control. Then, the proposed optimal control is validated by the industrial experiment, which the average oxygen and zinc oxide consumptions decreased by 568 m³/day (6.22%) and 3.09 t/day (5.16%) and the qualified rate of outlet ferrous ion concentrations increased by 5.5%, compared with the manual control.