Search

Discovering high utility sequences in a quantitative database is a popular data mining task. The goal is to enumerate all sequences of items (symbols) that have a high value for the user, as measured by a utility function. A representative application of high utility sequence mining is the identification of profitable sequences of purchases in transactions from online stores. Though useful, a drawback of that task is that the cost of items is not considered. However, cost is a key factor for decision-making in that domain and many others. To consider both the cost and utility of items for sequence mining, this paper defines a novel problem $$ \mathcal{FLCHUSM} $$ of mining frequent sequences having a high average utility and a low average cost. Though the proposed problem is a generalization of the traditional problem of frequent sequence mining, it is more challenging because the average utility and average cost functions do not satisfy the downward-closure property traditionally used to reduce the search space. To offer a solution to this issue, this paper presents a lower bound on the cost and two novel upper bounds on the utility. Besides, four width, depth pruning, reducing and tightening strategies are devised to eliminate unpromising patterns from the search space. Taking these theoretical results as a foundation, a new CUL (Cost-Utility List) data structure is conceived for storing and quickly updating the utility and cost information of patterns, and a novel algorithm named FLCHUSPM is proposed for $$ \mathcal{FLCHUSM} $$ . Results from several experiments show that FLCHUSPM is efficient in terms of memory usage and runtime, and that interesting patterns can be discovered in real data.

FPGAs, because of their energy efficiency, reconfigurability, and easily tunable HLS designs, have been used to accelerate an increasing number of machine learning, especially CNN-based, applications. As a representative example, IoT Edge applications, which require low latency processing of resource-hungry CNNs, offload the inferences from resource-limited IoT end nodes to Edge servers featuring FPGAs. However, the ever-increasing number of end nodes pressures these FPGA-based servers with new performance and adaptability challenges. While some works have exploited CNN optimizations to alleviate inferences’ computation and memory burdens, others have exploited HLS to tune accelerators for statically defined optimization goals. However, these works have not tackled both CNN and HLS optimizations altogether; neither have they provided any adaptability at runtime, where the workload’s characteristics are unpredictable. In this context, we propose a hybrid two-step approach that, first, creates new optimization opportunities at design-time through the automatic training of CNN model variants (obtained via pruning) and the automatic generation of versions of convolutional accelerators (obtained during HLS synthesis); and, second, synergistically exploits these created CNN and HLS optimization opportunities to deliver a fully dynamic Multi-FPGA system that adapts its resources in a fully automatic or user-configurable manner. We implement this two-step approach as the AdaServ Framework and show, through a smart video surveillance Edge application as a case study, that it adapts to the always-changing Edge conditions: AdaServ processes at least 3.37× more inferences (using the automatic approach) and is at least 6.68× more energy-efficient (user-configurable approach) than original convolutional accelerators and CNN Models (VGG-16 and AlexNet). We also show that AdaServ achieves better results than solutions dynamically changing only the CNN model or HLS version, highlighting the importance of exploring both; and that it is always better than the best statically chosen CNN model and HLS version, showing the need for dynamic adaptability.

Over the past several years, the need for on-device deep learning has been rapidly increasing, and efficient CNN inference on mobile platforms has been actively researched. Sparsity exploitation has been one of the most active research themes, but the studies mostly focus on weight sparsity by weight pruning. Activation sparsity, on the contrary, requires compression at runtime for every input tensor. Hence, the research on activation sparsity mainly targets NPUs that can efficiently process this with their own hardware logic. In this paper, we observe that it is difficult to accelerate CNN inference on mobile GPUs with natural activation sparsity and that the widely used CSR-based sparse convolution is not sufficiently effective due to the compression overhead. We propose several novel sparsification methods that can boost activation sparsity without harming accuracy. In particular, we selectively sparsify some layers with an extremely high sparsity and adopt sparse convolution or dense convolution depending on the layers. Further, we present an efficient sparse convolution method without compression and demonstrate that it can be faster than the CSR implementation. With ResNet-50, we achieved 1.88 speedup compared to TFLite on a Mali-G76 GPU.

The two objectives of early classification, accuracy and earliness, contradict with each other. In order to solve the problems of poor interpretation, huge candidate set of shapelets and adjustable quantification between the two objectives, a novel method of early classification of time series based on trend segmentation and optimization of cost function is proposed. Latent information of time series is mined by trend segmentation, and time stamp of discriminative shapelets is extracted. The number of shapelet candidates is greatly reduced by pruning based on the length and location, which improved the discrimination capability of chosen shapelets. An adjustable objective function is also defined to make a trade-off between accuracy and earliness, and then realize the early classification of time series. In view of the earliness and accuracy problems of different tendencies, this paper defines different coefficients to adjust the optimization objective function. The experimental results on UCR repository show that our proposed method achieves competitive results both at earliness and accuracy.

In target tracking WSNs, the tree-based and cluster-based network should be rebuilt when the location of target changes, which involves energy and time overheads. Although the chain structure solves this problem, it involves lots of unnecessary nodes and unacceptable transmitting delay. Therefore, we propose a dynamic chain-based collaboration (DCBC) approach for efficient target tracking and data gathering in this paper. DCBC completes the target tracking task by forming a dynamic tracking chain around the target. With the target moving, the dynamic tracking chain is adjusted by pruning some nodes as long as they are out of the range of the target and adding the nodes which entered the range of the target, so that the chain can adapt to the changing location of the target. The structure of the dynamic tracking chain does not need to be re-established when the target position changes. Furthermore, the sensing data are fused at every step when they are transmitted along the dynamic tracking chain. Using this way, the sensing data can be collected by tracking nodes and fused locally. In addition, we propose a dynamic and energy-efficient sleep scheduling for data transmission in the target tracking chain. The nodes in the tracking chain send their own data in turn. Thus it improves the energy utilization and data transmission efficiency of the network. The experimental results verify that DCBC can reduce and balance the average energy consumption of the network compared with the state-of-the-art approaches, and prolong the lifetime of networks.

Path planning is the basis and prerequisite for unmanned aerial vehicle (UAV) to perform tasks, and it is important to achieve precise location in path planning. This paper focuses on solving the UAV path planning problem under the constraint of system positioning error. Some nodes can re-initiate the accumulated flight error to zero and this type of scenario can be modeled as the resource-constrained shortest path problem with re-initialization (RCSPP-R). The additional re-initiation conditions expand the set of viable paths for the original constrained shortest path problem and increasing the search cost. To solve the problem, an effective preprocessing method is proposed to reduce the network nodes. At the same time, a relaxed pruning strategy is introduced into the traditional Pulse algorithm to reduce the search space and avoid more redundant calculations on unfavorable scalable nodes by the proposed heuristic search strategy. To evaluate the accuracy and effectiveness of the proposed algorithm, some numerical experiments were carried out. The results indicate that the three strategies can reduce the search space by 99%, 97% and 80%, respectively, and in the case of a large network, the heuristic algorithm combining the three strategies can improve the efficiency by an average of 80% compared to some classical solution.

Although deep convolutional neural networks (CNNs) have achieved significant success in computer vision applications, the real-world deployment of CNNs is often limited by computing resources and memory constraints. As a mainstream deep model compression technology, neural network pruning offers a promising prospect to reduce models’ parameters and calculation. In this paper, we proposed a novel filter pruning method that combines convolution filters and feature maps information for convolutional neural network compression, namely network pruning by using cluster similarity and large eigenvalues (CSHE). First, based on the convolution operation, we explore the similarity relationship of feature maps generated by the corresponding filters. Concretely, the clustering algorithm is used to classify the similarity of filter to guide the classification of feature map. Secondly, the proposed method utilizes the information of the large eigenvalues of the feature maps to rank the importance of filters. Finally, we prune the low-ranking filters and remain the high-ranking ones. The proposed method eliminates redundancy in convolution filters by applying large eigenvalues of feature maps based on filters similarity. In this way, most of the representative information in the network can be retained and the pruned results can be easily reproduced. Experiments show that the accuracy of the pruned sparse deep network obtained by the CSHE method in the classification tasks of CIFAR-10 and ImageNet ILSVRC-12 is almost the same as that of the reference network without any additional constraints.

Field-programmable Gate Array (FPGA) is a high-performance computing platform for Convolution Neural Networks (CNNs) inference. Winograd algorithm, weight pruning, and quantization are widely adopted to reduce the storage and arithmetic overhead of CNNs on FPGAs. Recent studies strive to prune the weights in the Winograd domain, however, resulting in irregular sparse patterns and leading to low parallelism and reduced utilization of resources. Besides, there are few works to discuss a suitable quantization scheme for Winograd. In this article, we propose a regular sparse pruning pattern in the Winograd-based CNN, namely, Sub-row-balanced Sparsity (SRBS) pattern, to overcome the challenge of the irregular sparse pattern. Then, we develop a two-step hardware co-optimization approach to improve the model accuracy using the SRBS pattern. Based on the pruned model, we implement a mixed precision quantization to further reduce the computational complexity of bit operations. Finally, we design an FPGA accelerator that takes both the advantage of the SRBS pattern to eliminate low-parallelism computation and the irregular memory accesses, as well as the mixed precision quantization to get a layer-wise bit width. Experimental results on VGG16/VGG-nagadomi with CIFAR-10 and ResNet-18/34/50 with ImageNet show up to 11.8×/8.67× and 8.17×/8.31×/10.6× speedup, 12.74×/9.19× and 8.75×/8.81×/11.1× energy efficiency improvement, respectively, compared with the state-of-the-art dense Winograd accelerator [20] with negligible loss of model accuracy. We also show that our design has 4.11× speedup compared with the state-of-the-art sparse Winograd accelerator [19] on VGG16.

Background DNA methylation and gene expression are known to play important roles in the etiology of human diseases such as prostate cancer (PCa). However, it has not yet been possible to incorporate information of DNA methylation and gene expression into polygenic risk scores (PRSs). Here, we aimed to develop and validate an improved PRS for PCa risk by incorporating genetically predicted gene expression and DNA methylation, and other genomic information using an integrative method. Methods Using data from the PRACTICAL consortium, we derived multiple sets of genetic scores, including those based on available single‐nucleotide polymorphisms through widely used methods of pruning and thresholding, LDpred, LDpred‐funt, AnnoPred, and EBPRS, as well as PRS constructed using the genetically predicted gene expression and DNA methylation through a revised pruning and thresholding strategy. In the tuning step, using the UK Biobank data (1458 prevalent cases and 1467 controls), we selected PRSs with the best performance. Using an independent set of data from the UK Biobank, we developed an integrative PRS combining information from individual scores. Furthermore, in the testing step, we tested the performance of the integrative PRS in another independent set of UK Biobank data of incident cases and controls. Results Our constructed PRS had improved performance (C statistics: 76.1%) over PRSs constructed by individual benchmark methods (from 69.6% to 74.7%). Furthermore, our new PRS had much higher risk assessment power than family history. The overall net reclassification improvement was 69.0% by adding PRS to the baseline model compared with 12.5% by adding family history. Conclusions We developed and validated a new PRS which may improve the utility in predicting the risk of developing PCa. Our innovative method can also be applied to other human diseases to improve risk prediction across multiple outcomes., An integrative score incorporating genetically predicted gene expression and DNA methylation and other genomic and non‐genomic information advanced our understanding of using genomic information to stratify subjects for prostate cancer.

This study presents a novel framework based on a heterogeneous ensemble method and a hybrid dimensionality reduction technique for spam detection in micro-blogging social networks. A hybrid of Information Gain (IG) and Principal Component Analysis (PCA) (dimensionality reduction) was implemented for the selection of important features and a heterogeneous ensemble consisting of Naïve Bayes (NB), K Nearest Neighbor (KNN), Logistic Regression (LR) and Repeated Incremental Pruning to Produce Error Reduction (RIPPER) classifiers based on Average of Probabilities (AOP) was used for spam detection. The proposed framework was applied on MPI_SWS and SAC’13 Tip spam datasets and the developed models were evaluated based on accuracy, precision, recall, f-measure, and area under the curve (AUC). From the experimental results, the proposed framework (that is, Ensemble + IG + PCA) outperformed other experimented methods on studied spam datasets. Specifically, the proposed method had an average accuracy value of 87.5%, an average precision score of 0.877, an average recall value of 0.845, an average F-measure value of 0.872 and an average AUC value of 0.943. Also, the proposed method had better performance than some existing methods. Consequently, this study has shown that addressing high dimensionality in spam datasets, in this case, a hybrid of IG and PCA with a heterogeneous ensemble method can produce a more effective method for detecting spam contents.

When using a Bayesian network to model a practical problem, weak prior knowledge projected as ancestral constraints is necessary. However, it is difficult to directly utilize these non-decomposable constraints using search strategies based on the decomposable score. In this study, we attempt to solve this problem by conducting an implicate path-space search graph and driving the A* algorithm, which is used to obtain the globally optimal solution satisfying the given constraints. We use a maximum covering principle to provide useful pruning rules based on these constraints in the new framework. Moreover, we improve the simple heuristic and the static k-cycle conflict heuristic to adapt to ancestral constraints. We theoretically prove that the new heuristic functions remain admissible and consistent. Our experiments demonstrate that the proposed framework with the new heuristic functions significantly reduces the space complexity of A* search compared with state-of-the-art frameworks, such as Bayesian network graphs and equivalent class trees, when integrating ancestral constraints.

The rapid development of autonomous driving in recent years presents many challenges for scene understanding. As an essential step towards scene understanding, semantic segmentation has received increased attention in the past few years. Although deep learning based approaches have achieved great success in improving the segmentation accuracy, most of them suffer from an inefficiency problem and can hardly be applied to real-time applications. In this paper, we analyze the computational cost of Convolutional Neural Network (CNN) and find that the inefficiency of CNNs is mainly caused by their wide structure rather than deep structure. In addition, the success of pruning based model compression methods proves that there are many redundant channels in CNNs. Thus, we design a narrow while deep backbone network to improve the efficiency of semantic segmentation. By casting our network to fully convolutional network (FCN32) segmentation architecture, the basic structure of most segmentation methods, we achieve 61.5% mIoU on Cityscapes validation dataset with only 4.2G floating-point operations (FLOPs) on $1024\times 2048$ inputs, which already outperforms one of the earliest real-time deep learning based segmentation methods: ENet (58.3% mIoU, 3.8G FLOPs on $640\times 360$ inputs). By further refining the output resolution of our network to the 1/8 of the input resolution with a simple encoder-decoder structure, we achieve 65.3% mIoU on Cityscapes test set with 14.0G FLOPs and 39.9 frames per second (FPS) on Titan X card. We have made our model publicly available at https://github.com/zgyang-hnu/NDNet .

Many emerging systems concurrently execute multiple applications that use deep neural network (DNN) as a key portion of the computation. To speedup the execution of such DNNs, various hardware accelerators have been proposed in recent works. Deep learning processor unit (DPU) from Xilinx is one such accelerator targeted for field programmable gate array (FPGA)-based systems. We study the runtime and energy consumption for different DNNs on a range of DPU configurations and derive useful insights. Using these insights, we formulate a design space exploration (DSE) strategy to explore tradeoffs in accuracy, runtime, cost, and energy consumption arising due to flexibility in choosing DNN topology, DPU configuration, and FPGA model. The proposed strategy provides a reduction of $28\times $ in the number of design points to be simulated and $23\times $ in the pruning time.

Evolving fuzzy systems (EFS) have received increased attention from the community for the purpose of data stream modeling in an incremental, single-pass and transparent manner. To date, a wide variety of EFS approaches have been developed and successfully used in real-world applications which address structural evolution and parameter adaptation in single EFS models. We propose a specific ensemble scheme of EFS to increase their robustness in predictive performance on new stream samples. Our approach relies on an online variant of bagging in which various EFS ensemble members are generated from online bags, that is, the members are updated based on a specific probabilistic online sampling technique, and this with guaranteed convergence to classical sampling in batch bagging. The autonomous pruning of ensemble members is undertaken to omit undesired members with atypically higher errors than other members. We propose two variants, hard pruning where undesired members are deleted forever from the ensemble, and soft pruning where members receive weights to calculate the overall ensemble prediction, according to their single performance; thus, members who are undesired at a certain point of time may be dynamically recalled at a later stage. The autonomous evolution of new ensemble members is carried out whenever a drift in the stream is detected, based on a significantly worsening performance indicator, measured in terms of the Hoeffding inequality. Newer members typically represent the drifted state better and are thus up-weighed compared to older members within an advanced (weighted) calculation of the overall ensemble prediction. The new approach termed online bagged EFS (OB-EFS) was successfully evaluated and compared with single EFS models and related SoA approaches on four data streams from real-world applications (containing various noise levels, drifts and new operating conditions) and showed significantly lower prediction error trend lines. Refereed/Peer-reviewed

Deep convolutional neural networks (CNNs) is difficult to deploy on resource-constrained devices due to its huge amount of computation. Channel pruning is an effective method to reduce the amount of computation and accelerate network inference. Most of channels pruning methods use statistics from a single structure (convolutional layer or batch normalization layer) of the sparse network to evaluate the importance of channels. The limitation of these methods is that it may often mistakenly delete the important channels. In view of this, we propose a novel method, namely Collaborative Channel Pruning (CCPrune), to evaluate the importance of channels, which combines the convolution layer weights and the BN layer scaling factors. The proposed method first introduces the regularization on the convolution layer weights and the BN layer scaling factors respectively. Then combine the weight of the convolutional layer and the scaling factor of the BN layer to evaluate the importance of the channel. Finally, it can delete the unimportant channels without reduces the performance of the model. The experimental results well demonstrate the effectiveness of our method. On CIFAR-10, it can reduce the FLOPs of VGG-19 by 85.50% while only slightly reducing the accuracy of the model, and it can reduce the FLOPs of Resnet-50 by 78.31% without reducing the accuracy of the model, respectively.

Broiler flock welfare is usually assessed through mortality, physiology, behavior, and walking ability. The possibility of assessing broiler chicken lameness using the bird walking ability was investigated using the machine learning approach for the first time. Data on broiler walking speed and acceleration, genetic strain, and sex were recorded and input in a dataset. Broilers were classified according to the 6-point gait score (GS0 is a sound bird, and GS5 is a severely lame bird). Decision trees were built initially using all datasets. The confusion matrix of each developed model was analyzed. The pruning technique was used, removing from the dataset the variables that did not infer in the classification results. We reorganized the dataset and re-arranged the data by grouping the intermediate target class of gait score using the Borda Count method. Re-processing data, we obtained a new set of decision trees. Using the 3-point gait score (GS0 is a sound bird, and GS2 is a lame bird), we obtained a new model with better accuracy (78%); however, the model had a lower accuracy for classifying lame broilers (GS2, 5%). The final decision tree was selected for classifying broilers, either sound or lame, according to their walking speed. The developed model presented good accuracy (91%), and it ordered properly sound (86%) and lame birds (92%). The novel model might be used to assess broiler lameness on-farm by registering the bird displacement velocity. Further developments using the model might allow flock lameness detection automatically.

Analyzing reachability in large discrete transition systems is an important sub-problem in several areas of AI, and of CS in general. State space search is a basic method for conducting such an analysis. A wealth of techniques have been proposed to reduce the search space without affecting the existence of (optimal) solution paths. In particular, strong stubborn set (SSS) pruning is a prominent such method, analyzing action dependencies to prune commutative parts of the search space. We herein show how to apply this idea to star-topology decoupled state space search, a recent search reformulation method invented in the context of classical AI planning. Star-topology decoupled state space search, short decoupled search, addresses planning tasks where a single center component interacts with several leaf components. The search exploits a form of conditional independence arising in this setting: given a fixed path p of transitions by the center, the possible leaf moves compliant with p are independent across the leaves. Decoupled search thus searches over center paths only, maintaining the compliant paths for each leaf separately. This avoids the enumeration of combined states across leaves. Just like standard search, decoupled search is adversely affected by commutative parts of its search space. The adaptation of strong stubborn set pruning is challenging due to the more complex structure of the search space, and the resulting ways in which action dependencies may affect the search. We spell out how to address this challenge, designing optimality-preserving decoupled strong stubborn set (DSSS) pruning methods. We introduce a design for star topologies in full generality, as well as simpler design variants for the practically relevant fork and inverted fork special cases. We show that there are cases where DSSS pruning is exponentially more effective than both, decoupled search and SSS pruning, exhibiting true synergy where the whole is more than the sum of its parts. Empirically, DSSS pruning reliably inherits the best of its components, and sometimes outperforms both.

Multiobjective optimization approaches have allowed the improvement of technical features in industrial processes, focusing on more accurate approaches for solving complex engineering problems and support decision-making. This paper proposes a hybrid approach to optimize the 3D printing technology parameters, integrating the design of experiments and multiobjective optimization methods, as an alternative to classical parametrization design used in machining processes. Alongside the approach, a multiobjective differential evolution with uniform spherical pruning (usp-MODE) algorithm is proposed to serve as an optimization tool. The parametrization design problem considered in this research has the following three objectives: to minimize both surface roughness and dimensional accuracy while maximizing the mechanical resistance of the prototype. A benchmark with non-dominated sorting genetic algorithm II (NSGA-II) and with the classical sp-MODE is used to evaluate the performance of the proposed algorithm. With the increasing complexity of engineering problems and advances in 3D printing technology, this study demonstrates the applicability of the proposed hybrid approach, finding optimal combinations for the machining process among conflicting objectives regardless of the number of decision variables and goals involved. To measure the performance and to compare the results of metaheuristics used in this study, three Pareto comparison metrics have been utilized to evaluate both the convergence and diversity of the obtained Pareto approximations for each algorithm: hyper-volume (H), g-Indicator (G), and inverted generational distance. To all of them, ups-MODE outperformed, with significant figures, the results reached by NSGA-II and sp-MODE algorithms.

Distance labeling approaches are widely adopted to speed up the online performance of shortest-distance queries. The construction of the distance labeling, however, can be exhaustive, especially on big graphs. For a major category of large graphs, small-world networks, the state-of-the-art approach is pruned landmark labeling ( $$\mathsf {PLL}$$ ). $${\mathsf {PLL}} $$ prunes distance labels based on a node order and directly constructs the pruned labels by performing breadth-first searches in the node order. The pruning technique, as well as the index construction, has a strong sequential nature which hinders $${\mathsf {PLL}} $$ from being parallelized. It becomes an urgent issue on massive small-world networks whose index can hardly be constructed by a single thread within a reasonable time. This paper first scales distance labeling on small-world networks by proposing a parallel shortest-distance labeling ( $$\mathsf {PSL}$$ ) scheme. $$\mathsf {PSL}$$ insightfully converts the $${\mathsf {PLL}} $$ ’s node-order dependency to a shortest-distance dependence, which leads to a propagation-based parallel labeling in D rounds where D denotes the diameter of the graph. To further scale up $$\mathsf {PSL}$$ , it is critical to reduce the index size. This paper proposes effective index compression techniques based on graph properties as well as label properties; it also explores best practices in using betweenness-based node order to reduce the index size. The efficient betweenness estimation of the graph nodes proposed may be of independent interest to graph practitioners. Extensive experimental results verify our efficiency on billion-scale graphs, near-linear speedup in a multi-core environment, and up to $$94\%$$ reduction in the index size.

To improve the performcance of community discovery algorithm applied to dynamic community detection objects, a parallel clustering analysis based on packet permission hierarchical association mining in community discovery of big data has been proposed. First, an evolutionary non-negative matrix decomposition framework based on clustering quality is proposed for dynamic community detection. Second, a clustering combined with dynamic pruning binary tree support vector machine (SVM) algorithm is proposed to prove the equivalence between evolutionary binary tree clustering and evolutionary module density optimization from the perspective of theoretical analysis. Based on this equivalence, a new semi-supervised association mining algorithm is proposed by adding prior information to the sample data without increasing the time complexity. Finally, through the experimental analysis on the static and dynamic community detection model, the performance advantage of the proposed algorithm on the community detection performance index is verified.

Top-k dominating (TKD) query is one of the methods to find the interesting objects by returning the k objects that dominate other objects in a given dataset. Incomplete datasets have missing values in uncertain dimensions, so it is difficult to obtain useful information with traditional data mining methods on complete data. BitMap Index Guided Algorithm (BIG) is a good choice for solving this problem. However, it is even harder to find top-k dominance objects on incomplete big data. When the dataset is too large, the requirements for the feasibility and performance of the algorithm will become very high. In this paper, we proposed an algorithm to apply MapReduce on the whole process with a pruning strategy, called Efficient Hadoop BitMap Index Guided Algorithm (EHBIG). This algorithm can realize TKD query on incomplete datasets through BitMap Index and use MapReduce architecture to make TKD query possible on large datasets. By using the pruning strategy, the runtime and memory usage are greatly reduced. What's more, we also proposed an improved version of EHBIG (denoted as IEHBIG) which optimizes the whole algorithm flow. Our in-depth work in this article culminates with some experimental results that clearly show that our proposed algorithm can perform well on TKD query in an incomplete large dataset and shows great performance in a Hadoop computing cluster.

Deep convolutional neural networks (DCNNs) have been successfully used in many computer visions task. However, with the increasing complexity of the network and continuous growth of data scale, training a DCNN model suffers from the following three problems: excessive network parameters, insufficient capability of the parameter optimization, and inefficient parallelism. To overcome these obstacles, this paper develops an optimization algorithm for deep convolutional neural network (FP-DCNN) in the MapReduce framework. First, a pruning method based on Taylor’s loss (FMPTL) is designed to trim redundant parameters, which not only compresses the structure of DCNN, but also reduces the computational cost of training. Next, a glowworm swarm optimization algorithm based on information sharing strategy (IFAS) is presented, which improves the ability of parameter optimization by adjusting the initialization of weights. Finally, a dynamic load balancing strategy based on parallel computing entropy (DLBPCE) is proposed to achieve an even distribution of data and thus improve the parallel performance of the cluster. Our experiments show that compared with other parallelized algorithms, this algorithm not only reduces the computational cost of network training, but also obtains a higher processing speed.

High-utility itemset mining (HUIM) is a useful tool for analyzing customer behavior in the field of data mining. HUIM algorithms can discover the most beneficial itemsets in transaction databases, namely the high-utility itemsets (HUIs), in contrast to frequent itemset mining (FIM) algorithms that rely on detecting frequent patterns. Several algorithms have been proposed to effectively carry out this task, but most of them ignore the categorization of items. In many real-world transaction databases, this helpful information about the categories and subcategories of items, represented as a taxonomy, is useful. Therefore, traditional HUIM algorithms can only discover itemsets at the lowest level of abstraction and leave out several important patterns from higher levels. To address this limitation, this work suggests the use of items taxonomy. Besides, to further enhance the performance of the task several effective pruning techniques are also revised and utilized to tighten the search space when considering the taxonomy of items. To accurately find multi-level HUIs from transaction databases enhanced with taxonomy information, a new algorithm called MLHMiner (Multiple-Level HMiner) is proposed, which is an extended version of the HMiner algorithm. We also prove that the pruning techniques of HMiner can be applied in different abstraction levels to efficiently mine multi-level HUIs. It can be seen from the experimental evaluations on several databases (both real and synthetic) that the designed approach is capable of identifying useful patterns from different abstraction levels with high efficiency.

Iron ore is an important raw material for the steel industry, so it is of great economic significance to determine the grade of the iron ore quickly and accurately. And the TFe content is the main indicator that determines the grade of the iron ore and whether the iron ore can be smelted directly. Unlike manual methods and methods for chemical analysis, the paper uses the selection of band for the near-infrared spectrum based on the pruning method and the two-hidden-layer extreme learning machine based on LU decomposition and seagull optimization algorithm (LU-TELM-SOA) to identify the TFe content. First of all, the paper proposes the selection of band based on the pruning method to retain the sensitive band of the near-infrared spectrum. Aiming at the problems of poor stability and low accuracy of a single LU-TELM (the two-hidden-layer extreme learning machine based on LU decomposition) model, the paper proposes LU-TELM-SOA. The experimental results show that LU-TELM-SOA has the advantages of high accuracy and strong stability.

In social networks, the strength of relationships between users can significantly affect the stability of the network. Two users are more likely to build the friendship if they share some common friends. Meanwhile, the breakdown or enhancement of critical connections may lead to a cascaded phenomenon and cause the social network collapsed or reinforced. In this paper, we leverage the k-truss model to measure the stability of a social network. To identify the critical edges, we propose two novel problems named k-truss minimization problem and k-truss maximization problem. Given a social network G, a positive integer k and a budget b, it aims to find b edges for deletion (resp. addition), which can lead to the maximum number of edges collapsed (resp. added) in the k-truss of G. We prove that both problems are NP-hard. To accelerate the computation, novel pruning rules and searching paradigms are developed for the corresponding problem. Comprehensive experiments are conducted over 9 real-life networks to demonstrate the effectiveness and efficiency of our proposed models and approaches.

In this article, we present a low-energy inference method for convolutional neural networks in image classification applications. The lower energy consumption is achieved by using a highly pruned (lower-energy) network if the resulting network can provide a correct output. More specifically, the proposed inference method makes use of two pruned neural networks (NNs), namely mildly and aggressively pruned networks, which are both designed offline. In the system, a third NN makes use of the input data for the online selection of the appropriate pruned network. The third network, for its feature extraction, employs the same convolutional layers as those of the aggressively pruned NN, thereby reducing the overhead of the online management. There is some accuracy loss induced by the proposed method where, for a given level of accuracy, the energy gain of the proposed method is considerably larger than the case of employing any one pruning level. The proposed method is independent of both the pruning method and the network architecture. The efficacy of the proposed inference method is assessed on Eyeriss hardware accelerator platform for some of the state-of-the-art NN architectures. Our studies show that this method may provide, on average, 70% energy reduction compared to the original NN at the cost of about 3% accuracy loss on the CIFAR-10 dataset.

Ensemble learning models have been widely used for wind power forecasting to facilitate efficient dispatching of power systems. However, traditional ensemble methods cannot always function well due to insufficient accuracy and diversity of base learners, ignorance of ensemble pruning, as well as the lack of adaptation capability. Therefore, a novel probabilistic wind power forecasting method is proposed based on selective ensemble of finite mixture Gaussian process regression models (SEFMGPR). First, a set of diverse local Gaussian process regression (GPR) models are constructed through multimodal perturbation mechanism, i.e., perturbing the training data and input attributes simultaneously. Then, a set of finite mixture GPR models (FMGPR) is built by integrating local GPR models through finite mixture mechanism (FMM). Next, the highly influential FMGPR models are selected using genetic algorithm (GA) based ensemble pruning. When a new test sample comes, the component predictions from the selected FMGPR models are adaptively combined by using FMM again and the probabilistic prediction results of the SEFMGPR model are obtained. Besides, an incremental adaptation mechanism is used to alleviate performance degradation of SEFMGPR. The application results from a real wind farm dataset show that SEFMGPR outperforms the traditional global and ensemble wind power prediction methods, and can maintain high prediction accuracy by effectively handling time-varying changes of wind power data.

While deep neural networks (DNNs) deliver state-of-the-art accuracy on various applications from face recognition to language translation, it comes at the cost of high computational and space complexity, hindering their deployment on edge devices. To enable efficient processing of DNNs in inference, a novel approach, called Evolutionary Multi-Objective Model Compression (EMOMC), is proposed to optimize energy efficiency (or model size) and accuracy simultaneously. Specifically, the network pruning and quantization space are explored and exploited by using architecture population evolution. Furthermore, by taking advantage of the orthogonality between pruning and quantization, a two-stage pruning and quantization co-optimization strategy is developed, which considerably reduces time cost of the architecture search. Lastly, different dataflow designs and parameter coding schemes are considered in the optimization process since they have a significant impact on energy consumption and the model size. Owing to the cooperation of the evolution between different architectures in the population, a set of compact DNNs that offer trade-offs on different objectives (e.g., accuracy, energy efficiency and model size) can be obtained in a single run. Unlike most existing approaches designed to reduce the size of weight parameters with no significant loss of accuracy, the proposed method aims to achieve a trade-off between desirable objectives, for meeting different requirements of various edge devices. Experimental results demonstrate that the proposed approach can obtain a diverse population of compact DNNs that are suitable for a broad range of different memory usage and energy consumption requirements. Under negligible accuracy loss, EMOMC improves the energy efficiency and model compression rate of VGG-16 on CIFAR-10 by a factor of more than 8 9. X and 2.4 X, respectively.

As the increasing availability and use of dynamic mobile communications, information from an Internet of Things (IoT) subset of devices, known as Internet of Connected Vehicles (IoCV), is collected with a level of uncertainty. To bridge this gap of data analytics, some studies take two factors individually to mine knowledge or information, such as uncertainty and utility as two exemplary factors. However, this approach may cause actual loss of knowledge integrity. In this work, our first result is a knowledge called H igh E xpected U tility S equential P atterns (HEUSPs) that is both novel and also provides an alternative option for knowledge discovery regarding utility and uncertainty factors by a single threshold in IoCV environments. Furthermore, two PUL-Chain and EUL-Chain structures with six pruning methodologies are respectively developed to maintain information that is necessary and reduce the search space for improving mining performance. Our experimental results show both efficiency and strength of the designed algorithm compared to HUS-Span which is considered to be the current standard in utility-oriented sequential pattern mining.

The time and monetary costs of training sophisticated deep neural networks are exorbitant, which motivates resource-limited users to outsource the training process to the cloud. Concerning that an untrustworthy cloud service provider may inject backdoors to the returned model, the user can leverage state-of-the-art defense strategies to examine the model. In this paper, we aim to develop robust backdoor attacks (named RobNet) that can evade existing defense strategies from the standpoint of malicious cloud providers. The key rationale is to diversify the triggers and strengthen the model structure so that the backdoor is hard to be detected or removed. To attain this objective, we refine the trigger generation algorithm by selecting the neuron(s) with large weights and activations and then computing the triggers via gradient descent to maximize the value of the selected neuron(s). In stark contrast to existing works that fix the trigger location, we design a multi-location patching method to make the model less sensitive to mild displacement of triggers in real attacks. Furthermore, we extend the attack space by proposing multi-trigger backdoor attacks that can misclassify inputs with different triggers into the same or different target label(s). We evaluate the performance of RobNet on MNIST, GTSRB, and CIFAR-10 datasets, against four representative defense strategies Pruning, NeuralCleanse, Strip, and ABS. The comparison with two state-of-the-art baselines BadNets and Hidden Backdoors demonstrates that RobNet achieves higher attack success rate and is more resistant to potential defenses.

An important data mining task is to discover all high utility sequences in a quantitative sequence database. Although useful, the number of discovered sequences is often very large. To find patterns that are more tailored to a user’s needs, this paper studies the problem of mining frequent high utility sequences satisfying item constraints. This article proposes a novel algorithm named C-FHUSM to quickly obtain these sequences from two concise representations discovered from a quantitative sequence database, namely frequent generator high utility sequences and frequent closed high utility sequences. The first set is extracted using a novel algorithm named FGenHUSM, while an existing algorithm is applied to extract the second set. C-FHUSM integrates novel pruning techniques to ignore sequences that do not satisfy item constraints early by checking only a small number of representative sequences at the beginning of the mining process. Experimental results show that C-FHUSM can be more than ten times faster and has better scalability than a modified version of the state-of-the-art EHUSM algorithm for mining sequences with item constraints. Moreover, it is found that using C-FHUSM is beneficial when a user frequently changes constraints as results can be updated without rescanning the database.

High Utility Itemset Mining (HUIM) is one of the most investigated tasks of data mining. It has broad applications in domains such as product recommendation, market basket analysis, e-learning, text mining, bioinformatics, and web click stream analysis. Insights from such pattern analysis provide numerous benefits, including cost cutting, improved competitive advantage, and increased revenue. However, HUIM methods may discover misleading patterns as they do not evaluate the correlation of extracted patterns. As a consequence, a number of algorithms have been proposed to mine correlated HUIs. These algorithms still suffer from the issue of the computational cost in terms of both time and memory consumption. This paper presents an algorithm, named Efficient Correlated High Utility Pattern Mining (ECoHUPM), to efficiently mine the high utility patterns having strong correlation items. A new data structure based on utility tree (UTtree) named CoUTlist is proposed to store sufficient information for mining the desired patterns. Three pruning properties are introduced to reduce the search space and improve the mining performance. Experiments on sparse, very sparse, dense, and very dense datasets indicate that the proposed ECoHUPM algorithm is efficient as compared to the state-of-the-art CoHUIM and CoHUI-Miner algorithms in terms of both time and memory consumption.

In intrusion detection, approaches incorporated with data mining become interesting nowadays, in particular the associative classification which is a hybrid technique which uses pruning measure. Although the full rules set is not intended for precise classification, the rules have been used effectively by classifiers that have been built in previous systems. Class detection by variance process uses the association rule mining concept for discovering the association among data variables, and the gained information about different patterns is used to classify the variables into different classes. Using the identified centroids, numerical data is discretized and fed to rule precision index (RPI) classifier for rule induction. Popular Data mining tools operate this technique in the name association based on classification (CBA) which uses confidence as an interest measure for rule pruning. In this work, we present a new interest measure named rule precision index (RPI) which helps us to prune association rules efficiently, and the impact is observed in the classification of attack and non-attack. The resultant associate method produces the best performance among association-based classifiers and is evaluated with conventional classifiers against three different intrusion detection datasets, namely NSL-KDD, CICIDS-2017 and KDD CUP99. The proposed RPI classifier, incorporated with novel interest measure, provides the best accuracy rate of 89.48 % on average than the available classifiers.

One of the biggest open questions in the field of asteroid mining are the unknown characteristics and composition of more than 99% of the discovered objects. In order to minimize the risks of false investment, this uncertainty has to be diminished. One strategy to overcome this is to realize space missions that characterize the composition of asteroids in an efficient way. The above described issue is addressed in this paper. A target accessibility analysis is presented, that is tailored for asteroid characterization missions departing from a main belt parking orbit. To enable and facilitate a flexible, non predefined and autonomous object selection, adequate database constraints based on the orbital elements of the targets at departure are presented that enable the filtering of non-reachable objects without excluding potential targets. The constraints are applied to all asteroids currently listed in the JPL Small-Body Database Browser. With the derived constraints, the used database can be reduced to 1% of its original size.

Temporal graphs are equipped with entities and the relationships between entities associated with time stamps. Cohesive subgraph mining (CSM) is a fundamental task in temporal graph analysis, which has gathered great research interests. It benefits from reflecting the dynamism of graphs and has many real-world applications. Yet, most existing work focus on the cohesive subgraph detection (CSD) problem, which identifies all the defined subgraphs in the entire temporal graphs. When graph size becoming too large, it is impractical. In this paper, we are the first to concern about the cohesive subgraph search (CSS) problem in large temporal graphs. In specific, given a query vertex, we are seeking the continual densely connected subgraph including the query vertex. To this end, (1) we model the cohesive subgraph in temporal graphs as a (𝜃,τ)-continual k-core and prove its NP-hardness; (2) we develop two exact algorithms based on different vertex enumeration strategies, called Exact-VD and Exact-VE, respectively. Exact-VD uses depth-first search to find the target subgraphs in a top-down way by gradually deleting vertices from the current subgraph; while Exact-VE starts from the query vertex and continuously expands the ranked vertices in the candidate group until reaching the target subgraphs. Meanwhile, several elegant pruning rules are designed to reduce the search space; (3) to further speed up, we propose an efficient approximate local search method, called Approx-LS, which greedily expands the current subgraph guided by the developed heuristic functions until identifying the results. Comprehensive experiments on four real-life datasets verify the efficiency and effectiveness of our proposed approaches.

Multimedia Internet-of-Things (IoT) systems have been widely utilized in various computer vision tasks and significantly integrated computer vision and networking capabilities. In these systems, convolutional neural networks (CNNs) perform a preliminary analysis of the collected video or image information in the edge devices. However, the high computational cost and huge storage consumption of the complex CNNs prevent their deployment on mobile-edge devices that have limited computational resource and memory. In this article, we aim to simultaneously accelerate and compress CNNs via a multilevel filter pruning (MFP) algorithm, to alleviate the dependence on the hardware of IoT edge nodes. First, a global pruning sensitivity order is defined, which could guide us to perform preliminary pruning from the perspective of convolutional layers’ sensitivity. Then, the functional index of each filter is judged by the image entropy of its output feature map, which contributes to further pruning from the perspective of filter function importance. Finally, the moderate fine tuning is adopted to recover the network capability. The experimental results show that the proposed MFP algorithm could reduce 54.5% floating-point operations and 31.9% graphics memory for VGG-16 on CIFAR-10, and achieve $5.45 \times $ floating-point acceleration and $19.70 \times $ storage reduction for VGG-16 on ImageNet. In the reconstruction phase, the algorithm could recover the network capability much faster than the existing pruning algorithms.

Currently, with the continuous development of relation extraction tasks, we notice that the ability to extract nontaxonomic relations has improved frustratingly slowly, and the only relation extraction dataset in the field of public opinion is the New York Times dataset (NYT) annotated by distant supervision. This paper simultaneously addresses two issues. We first propose a new model that is tailored for nontaxonomic relation extraction, which combines a context-aware model with a weighted graph convolutional network (WGCN) model characterized by dependency trees. It effectively blends contextual and dependent structural information. We further apply a pruning strategy to the input tree so that the model can effectively retain valid information and delete redundant information. Then, we build a supervised Chinese relation extraction dataset, XUNRED (Xinjiang University Nontaxonomic Relation Extraction Dataset), which is obtained after manually tagging the Baidu Encyclopedia, Baidu Post Bar and Baidu Information Flow text, and address the nontaxonomic relation in the public opinion domain. The experimental results on this new dataset show that our model can combine the contextual information with the structural information in the dependency tree better than other models.

Given a large collection of transactions containing items, a basic common association rules problem is the huge size of the extracted rule set. Pruning uninteresting and redundant association rules is a promising approach to solve this problem. In this paper, we propose a Condensed Representation for Positive and Negative Association Rules representing non-redundant rules for both exact and approximate association rules based on the sets of frequent generator itemsets, frequent closed itemsets, maximal frequent itemsets, and minimal infrequent itemsets in database B. Experiments on dense (highly-correlated) databases show a significant reduction of the size of extracted association rule set in database B.

Despite the outstanding performance, deep convolutional neural networks (CNNs) are computationally expensive and contain a large number of redundant parameters, hindering their deployment on resource constrained platforms. To address this issue, many model compression methods have been proposed. However, these methods mainly focus on pruning redundant parameters or designing efficient architectures, the redundancy in convolution kernels has rarely been investigated. In this paper, we find that the contributions of parameters at different locations in the traditional 3 × 3 kernels are not the same, and this distribution varies considerably in different layers. Motivated by this, we propose to use irregular kernels and present a novel approach to implementing convolution with any-shape kernels (ASKs) efficiently. The proposed ASKs are plug-and-play and can be readily embedded into existing CNNs, providing efficient modules for building compact CNNs. Experiments on benchmarks demonstrate the effectiveness of the proposed method. We improve the accuracy of VGG-16 on CIFAR-10 dataset from 93.45% to 94.04% simply by replacing the regular 3 × 3 kernel with cross-shaped kernel, which takes up only about 5 / 9 of the original storage and computing resources. Compared to state-of-the-art model compression methods, our ASKs achieve a better trade-off between accuracy and compression ratio.

Skyline computation, aiming at identifying a set of skyline points that are not dominated by any other point, is particularly useful for multi-criteria data analysis and decision making. Traditional skyline computation, however, is inadequate to answer queries that need to analyze not only individual points but also groups of points. To address this gap, we generalize the original skyline definition to the novel group-based skyline (G-Skyline), which represents Pareto optimal groups that are not dominated by other groups. In order to compute G-Skyline groups consisting of $s$ s points efficiently, we present a novel structure that represents the points in a directed skyline graph and captures the dominance relationships among the points based on the first $s$ s skyline layers. We propose efficient algorithms to compute the first $s$ s skyline layers. We then present two heuristic algorithms to efficiently compute the G-Skyline groups: the point-wise algorithm and the unit group-wise algorithm, using various pruning strategies. We observe that the number of G-Skyline groups of a dataset can be significantly large, we further propose the top- $k$ k representative G-Skyline groups based on the number of dominated points and the number of dominated groups and present efficient algorithms for computing them. The experimental results on the real NBA dataset and the synthetic datasets show that G-Skyline is interesting and useful, and our algorithms are efficient and scalable.

Dynamic activation pruning of convolutional neural networks (CNNs) is a class of techniques that reduce both runtime and memory usage in CNN implementations by skipping unnecessary or low-impact computations in convolutional layers. However, since dynamic pruning results in different sequences of memory accesses depending on the input to the CNN, they potentially open the door to inference-phase side-channel attacks that may leak private data with each input. We demonstrate a memory-based attack inferring a dynamically-pruned CNN’s outputs for various victim CNN models and datasets. We find that an attacker can train their own machine learning model to learn to guess victim image classifications using the victim’s memory access patterns with significantly better than random chance. Moreover, unlike previous related work, our attack: 1) continually leaks user data for each input and 2) does not require adversarial presence during the victim training.

In this letter, we consider the decoding delay minimizing problem for delivering a frame of packets to a set of user-devices (UDs) using instantly decodable network coding (IDNC). In the considered device-to-device (D2D) network, UDs have limited coverage zones that represent clusters and can speed up the delivery of the requested packets of other UDs by sending IDNC packets. The decoding delay minimization problem is a joint optimization problem of selecting the transmitting UDs and their coding decisions. In this work, we propose a low complexity, yet optimal, solution to the decoding delay minimization problem using graph pruning method. Our proposed innovative method introduces a sequential pruning algorithm that judiciously generates clusters that are certainly contributing to the network while simultaneously designing a new multi-layer IDNC graph. We also prove that the optimal solution to the problem can be achieved by the generated clusters by our proposed algorithm. Numerical results reveal that the proposed solution significantly reduces the computational complexity compared to the existing method with similar decoding delay performance.

In this letter, we apply the model compression, i.e., network pruning, into wireless federated learning (FL) system to mitigate the local computation and communication bottlenecks. Firstly, the convergence rate and learning latency of the FL system are mathematically analyzed. Then, an optimization problem is formulated to maximize the convergence rate while guaranteeing the learning latency via jointly optimizing the pruning ratio and spectrum allocation. Finally, the experimental results show that the proposed learning scheme can improve the performance of the wireless FL as compared with other conventional schemes.

Uncertain data management has recently attracted much research interest in the networking community and database community, as in many emerging applications, e.g., sensor data monitoring, location-based services and vehicle tracking, where data are inherently uncertain due to measurement errors, update delays etc. The probabilistic k nearest neighbor (k-PNN) query returns k objects with the highest probability of being the k-th nearest neighbor from a given query point Q. However, compared to the traditional k-NN over precise data, the computation cost of k-PNN is rather expensive due to the costly numerical integration or Monte-Carlo approach it adopts, which raises a challenge in answering k-PNN in very large uncertain sensor networks. To address this challenge, we propose an efficient strategy for processing k-PNN queries. Specifically, we introduce two effective pruning methods, spatial pruning and probabilistic pruning, to speed up the query procedure by reducing the search space. The spatial pruning is based on the bounding region of the k-th nearest neighbor, while the probabilistic pruning is based on the lower and upper probability bounds of each k-NN candidate object after spatial pruning. Extensive experiments have been implemented to demonstrate the efficiency and effectiveness of our proposed method under various settings, in terms of both wall clock time and the number of candidate objects to be evaluated.

Applying Compressive Sensing (CS) to Received Signal Strength (RSS) based multi-emitter localization using Unmanned Arial Vehicles (UAVs) attracts much attention for its simplicity and efficiency. However, the RSS-based CS approach is vulnerable to the noise in a practical scenario. To mitigate this, we propose a robust localization framework for multiple emitters in UAV-based Wireless Sensor Network (WSN). We first approximate the lognormal noise’s influence on the dictionary by a two-layer hierarchical prior model. Then, by exploiting multi-frequency measurements, the multi-emitter localization is transformed into the joint estimation for multiple sparse vectors and noise level. Finally, the joint estimation problem is solved by a Concurrent Variational Bayesian Inference (CVBI) algorithm, where an adaptive grid pruning mechanism is designed. The merits of the proposed framework are testified by numerical simulations.

Community search in node-attributed network returns a subgraph that contains the query vertices, and can be used in a wide range of applications domain in real world. However, most existing works ignore the influence among nodes. This leads to inaccuracy in many scenarios which require communities with high expansibility. So a few recent studies have tried to consider the influence in community search. But they ignore the attributes cohesiveness of community. Therefore, this research proposes two influential community search algorithms by taking both the influence and node attributes into consideration. We propose a novel scoring function to check the cohesiveness of the community. Our method can find the attributed pkd-truss community by maximizing the attribute and influence relevance scoring function. In order to enable effective community search in large network, we develop a graph refining algorithm and pruning rule. So our methods can provide more personalized and effective subgraph searching for big data, which can enable many downstream applications such as recommendation of dating sites, setting up of social events and other e-commerce applications. The experimental work on four networks with ground-truth communities confirms that our methods can effectively find the personalized community and have better performance than existing approaches.

Explicit model predictive control is an established methodology for the offline determination of the optimal control policy for linear discrete time-invariant systems with linear constraints. Nevertheless, nonlinearities in the form of quadratic constraints naturally appear in process models or are imposed for stability purposes in model predictive control formulations. In this manuscript, we present the theoretical developments and propose an algorithm for the exact solution of explicit nonlinear model predictive control problems with convex quadratic constraints. Our approach is based on a second-order Taylor approximation of Fiacco’s Basic Sensitivity Theorem, which allows for the existence and the analytic derivation of the optimal control actions. The complete exploration of the parameter space is founded on an active set strategy, which employs a pruning criterion to eliminate infeasible active sets. Based on that, the optimal map of solutions is constructed along with the corresponding control actions. The proposed strategy is applied to an explicit nonlinear model predictive control problem with an ellipsoidal terminal set, and comparisons with approximate solutions are drawn to demonstrate the benefits of the presented approach. Furthermore, as a practical application, the optimal operation of a chemostat in the presence of disturbances is exhibited.