Classical Benchmark Research Articles

MetaCARD: Meta-Reinforcement Learning with Task Uncertainty Feedback via Decoupled Context-Aware Reward and Dynamics Components

Meta-Reinforcement Learning (Meta-RL) aims to reveal shared characteristics in dynamics and reward functions across diverse training tasks. This objective is achieved by meta-learning a policy that is conditioned on task representations with encoded trajectory data or context, thus allowing rapid adaptation to new tasks from a known task distribution. However, since the trajectory data generated by the policy may be biased, the task inference module tends to form spurious correlations between trajectory data and specific tasks, thereby leading to poor adaptation to new tasks. To address this issue, we propose the Meta-RL with task unCertAinty feedback through decoupled context-aware Reward and Dynamics components (MetaCARD). MetaCARD distinctly decouples the dynamics and rewards when inferring tasks and integrates task uncertainty feedback from policy evaluation into the task inference module. This design effectively reduces uncertainty in tasks with changes in dynamics or/and reward functions, thereby enabling accurate task identification and adaptation. The experiment results on both Meta-World and classical MuJoCo benchmarks show that MetaCARD significantly outperforms prevailing Meta-RL baselines, demonstrating its remarkable adaptation ability in sophisticated environments that involve changes in both reward functions and dynamics.

Unsupervised Layer-Wise Score Aggregation for Textual OOD Detection

Out-of-distribution (OOD) detection is a rapidly growing field due to new robustness and security requirements driven by an increased number of AI-based systems. Existing OOD textual detectors often rely on anomaly scores (\textit{e.g.}, Mahalanobis distance) computed on the embedding output of the last layer of the encoder. In this work, we observe that OOD detection performance varies greatly depending on the task and layer output. More importantly, we show that the usual choice (the last layer) is rarely the best one for OOD detection and that far better results can be achieved, provided that an oracle selects the best layer. We propose a data-driven, unsupervised method to leverage this observation to combine layer-wise anomaly scores. In addition, we extend classical textual OOD benchmarks by including classification tasks with a more significant number of classes (up to 150), which reflects more realistic settings. On this augmented benchmark, we show that the proposed post-aggregation methods achieve robust and consistent results comparable to using the best layer according to an oracle while removing manual feature selection altogether.

Robust Uncertainty Quantification Using Conformalised Monte Carlo Prediction

Deploying deep learning models in safety-critical applications remains a very challenging task, mandating the provision of assurances for the dependable operation of these models. Uncertainty quantification (UQ) methods estimate the model’s confidence per prediction, informing decision-making by considering the effect of randomness and model misspecification. Despite the advances of state-of-the-art UQ methods, they are computationally expensive or produce conservative prediction sets/intervals. We introduce MC-CP, a novel hybrid UQ method that combines a new adaptive Monte Carlo (MC) dropout method with conformal prediction (CP). MC-CP adaptively modulates the traditional MC dropout at runtime to save memory and computation resources, enabling predictions to be consumed by CP, yielding robust prediction sets/intervals. Throughout comprehensive experiments, we show that MC-CP delivers significant improvements over comparable UQ methods, like MC dropout, RAPS and CQR, both in classification and regression benchmarks. MC-CP can be easily added to existing models, making its deployment simple. The MC-CP code and replication package is available at https://github.com/team-daniel/MC-CP.

Dual-Level Curriculum Meta-Learning for Noisy Few-Shot Learning Tasks

Few-shot learning (FSL) is essential in many practical applications. However, the limited training examples make the models more vulnerable to label noise, which can lead to poor generalization capability. To address this critical challenge, we propose a curriculum meta-learning model that employs a novel dual-level class-example sampling strategy to create a robust curriculum for adaptive task distribution formulation and robust model training. The dual-level framework proposes a heuristic class sampling criterion that measures pairwise class boundary complexity to form a class curriculum; it uses effective example sampling through an under-trained proxy model to form an example curriculum. By utilizing both class-level and example-level information, our approach is more robust to handle limited training data and noisy labels that commonly occur in few-shot learning tasks. The model has efficient convergence behavior, which is verified through rigorous convergence analysis. Additionally, we establish a novel error bound through a hierarchical PAC-Bayesian analysis for curriculum meta-learning under noise. We conduct extensive experiments that demonstrate the effectiveness of our framework in outperforming existing noisy few-shot learning methods under various few-shot classification benchmarks. Our code is available at https://github.com/ritmininglab/DCML.

GradTree: Learning Axis-Aligned Decision Trees with Gradient Descent

Decision Trees (DTs) are commonly used for many machine learning tasks due to their high degree of interpretability. However, learning a DT from data is a difficult optimization problem, as it is non-convex and non-differentiable. Therefore, common approaches learn DTs using a greedy growth algorithm that minimizes the impurity locally at each internal node. Unfortunately, this greedy procedure can lead to inaccurate trees. In this paper, we present a novel approach for learning hard, axis-aligned DTs with gradient descent. The proposed method uses backpropagation with a straight-through operator on a dense DT representation, to jointly optimize all tree parameters. Our approach outperforms existing methods on binary classification benchmarks and achieves competitive results for multi-class tasks. The implementation is available under: https://github.com/s-marton/GradTree

Arbitrariness and Social Prediction: The Confounding Role of Variance in Fair Classification

Variance in predictions across different trained models is a significant, under-explored source of error in fair binary classification. In practice, the variance on some data examples is so large that decisions can be effectively arbitrary. To investigate this problem, we take an experimental approach and make four overarching contributions. We: 1) Define a metric called self-consistency, derived from variance, which we use as a proxy for measuring and reducing arbitrariness; 2) Develop an ensembling algorithm that abstains from classification when a prediction would be arbitrary; 3) Conduct the largest to-date empirical study of the role of variance (vis-a-vis self-consistency and arbitrariness) in fair binary classification; and, 4) Release a toolkit that makes the US Home Mortgage Disclosure Act (HMDA) datasets easily usable for future research. Altogether, our experiments reveal shocking insights about the reliability of conclusions on benchmark datasets. Most fair binary classification benchmarks are close-to-fair when taking into account the amount of arbitrariness present in predictions -- before we even try to apply any fairness interventions. This finding calls into question the practical utility of common algorithmic fairness methods, and in turn suggests that we should reconsider how we choose to measure fairness in binary classification.

Decoupled Contrastive Learning for Long-Tailed Recognition

Supervised Contrastive Loss (SCL) is popular in visual representation learning. Given an anchor image, SCL pulls two types of positive samples, i.e., its augmentation and other images from the same class together, while pushes negative images apart to optimize the learned embedding. In the scenario of long-tailed recognition, where the number of samples in each class is imbalanced, treating two types of positive samples equally leads to the biased optimization for intra-category distance. In addition, similarity relationship among negative samples, that are ignored by SCL, also presents meaningful semantic cues. To improve the performance on long-tailed recognition, this paper addresses those two issues of SCL by decoupling the training objective. Specifically, it decouples two types of positives in SCL and optimizes their relations toward different objectives to alleviate the influence of the imbalanced dataset. We further propose a patch-based self distillation to transfer knowledge from head to tail classes to relieve the under-representation of tail classes. It uses patch-based features to mine shared visual patterns among different instances and leverages a self distillation procedure to transfer such knowledge. Experiments on different long-tailed classification benchmarks demonstrate the superiority of our method. For instance, it achieves the 57.7% top-1 accuracy on the ImageNet-LT dataset. Combined with the ensemble-based method, the performance can be further boosted to 59.7%, which substantially outperforms many recent works. Our code will be released.

Curriculum-Enhanced Residual Soft An-Isotropic Normalization for Over-Smoothness in Deep GNNs

Despite Graph neural networks' significant performance gain over many classic techniques in various graph-related downstream tasks, their successes are restricted in shallow models due to over-smoothness and the difficulties of optimizations among many other issues. In this paper, to alleviate the over-smoothing issue, we propose a soft graph normalization method to preserve the diversities of node embeddings and prevent indiscrimination due to possible over-closeness. Combined with residual connections, we analyze the reason why the method can effectively capture the knowledge in both input graph structures and node features even with deep networks. Additionally, inspired by Curriculum Learning that learns easy examples before the hard ones, we propose a novel label-smoothing-based learning framework to enhance the optimization of deep GNNs, which iteratively smooths labels in an auxiliary graph and constructs many gradual non-smooth tasks for extracting increasingly complex knowledge and gradually discriminating nodes from coarse to fine. The method arguably reduces the risk of overfitting and generalizes better results. Finally, extensive experiments are carried out to demonstrate the effectiveness and potential of the proposed model and learning framework through comparison with twelve existing baselines including the state-of-the-art methods on twelve real-world node classification benchmarks.

ReGCL: Rethinking Message Passing in Graph Contrastive Learning

Graph contrastive learning (GCL) has demonstrated remarkable efficacy in graph representation learning. However, previous studies have overlooked the inherent conflict that arises when employing graph neural networks (GNNs) as encoders for node-level contrastive learning. This conflict pertains to the partial incongruity between the feature aggregation mechanism of graph neural networks and the embedding distinction characteristic of contrastive learning. Theoretically, to investigate the location and extent of the conflict, we analyze the participation of message-passing from the gradient perspective of InfoNCE loss. Different from contrastive learning in other domains, the conflict in GCL arises due to the presence of certain samples that contribute to both the gradients of positive and negative simultaneously under the manner of message passing, which are opposite optimization directions. To further address the conflict issue, we propose a practical framework called ReGCL, which utilizes theoretical findings of GCL gradients to effectively improve graph contrastive learning. Specifically, two gradient-based strategies are devised in terms of both message passing and loss function to mitigate the conflict. Firstly, a gradient-guided structure learning method is proposed in order to acquire a structure that is adapted to contrastive learning principles. Secondly, a gradient-weighted InfoNCE loss function is designed to reduce the impact of false negative samples with high probabilities, specifically from the standpoint of the graph encoder. Extensive experiments demonstrate the superiority of the proposed method in comparison to state-of-the-art baselines across various node classification benchmarks.

Learning to Unlearn: Instance-Wise Unlearning for Pre-trained Classifiers

Since the recent advent of regulations for data protection (e.g., the General Data Protection Regulation), there has been increasing demand in deleting information learned from sensitive data in pre-trained models without retraining from scratch. The inherent vulnerability of neural networks towards adversarial attacks and unfairness also calls for a robust method to remove or correct information in an instance-wise fashion, while retaining the predictive performance across remaining data. To this end, we consider instance-wise unlearning, of which the goal is to delete information on a set of instances from a pre-trained model, by either misclassifying each instance away from its original prediction or relabeling the instance to a different label. We also propose two methods that reduce forgetting on the remaining data: 1) utilizing adversarial examples to overcome forgetting at the representation-level and 2) leveraging weight importance metrics to pinpoint network parameters guilty of propagating unwanted information. Both methods only require the pre-trained model and data instances to forget, allowing painless application to real-life settings where the entire training set is unavailable. Through extensive experimentation on various image classification benchmarks, we show that our approach effectively preserves knowledge of remaining data while unlearning given instances in both single-task and continual unlearning scenarios.

Performance of dental students, orthodontic residents, and orthodontists for classification of midpalatal suture maturation stages on cone-beam computed tomography scans – a preliminary study

BackgroundAssessment of midpalatal suture maturation on cone-beam computed tomography (CBCT) scans is performed by visual inspection and is therefore subjective. The extent to which the assessment of midpalatal suture maturation is affected by rater experience has not been adequately explored in the existing literature, thus limiting the availability of evidence-based findings. This study compared the outcomes of classification by dental students, orthodontic residents, and orthodontists.MethodsThree different groups of students, orthodontic residents, and orthodontists evaluated 10 randomly chosen CBCT scans regarding midpalatal suture maturation from a pool of 179 patients (98 female and 81 male patients) aged 8 – 40 years which were previously classified by evaluating CBCT scans. The pool was set as benchmark utilizing midpalatal suture maturation classification by one examiner (OsiriX Lite version 11.0; Pixmeo SARL, Bernex, Switzerland). For assessment of intra-rater reliability of the examiners of each group the randomly chosen subjects were reclassified for midpalatal suture maturation after a wash-out period of two weeks by using the same software. Statistical analysis was performed to evaluate intra- and interrater reliability of the three groups with differing experience level.ResultsGroupwise intra-rater reliability assessment between the classification and reclassification was weak for examiners with a low level of experience (k = 0.59). Orthodontists had highest degree of agreement with regard to benchmark classification with an inter-rater reliability to be considered as moderate (k = 0.68).ConclusionsAssessment of midpalatal suture maturation on CBCT scans appears to be a subjective process and is considerably related to the experience level of the examiner. A high level of clinical experience seems to be favorable but does not necessarily ensure accurate results.

Adaptive habitat biogeography-based optimizer for optimizing deep CNN hyperparameters in image classification

Deep Convolutional Neural Networks (DCNNs) have shown remarkable success in image classification tasks, but optimizing their hyperparameters can be challenging due to their complex structure. This paper develops the Adaptive Habitat Biogeography-Based Optimizer (AHBBO) for tuning the hyperparameters of DCNNs in image classification tasks. In complicated optimization problems, the BBO suffers from premature convergence and insufficient exploration. In this regard, an adaptable habitat is presented as a solution to these problems; it would permit variable habitat sizes and regulated mutation. Better optimization performance and a greater chance of finding high-quality solutions across a wide range of problem domains are the results of this modification's increased exploration and population diversity. AHBBO is tested on 53 benchmark optimization functions and demonstrates its effectiveness in improving initial stochastic solutions and converging faster to the optimum. Furthermore, DCNN-AHBBO is compared to 23 well-known image classifiers on nine challenging image classification problems and shows superior performance in reducing the error rate by up to 5.14%. Our proposed algorithm outperforms 13 benchmark classifiers in 87 out of 95 evaluations, providing a high-performance and reliable solution for optimizing DNNs in image classification tasks. This research contributes to the field of deep learning by proposing a new optimization algorithm that can improve the efficiency of deep neural networks in image classification.

An ensemble method with a hybrid of genetic algorithm and K-prototypes algorithm for mixed data classification

Due to challenges posed by mixed data clustering, this study aims to introduce an innovative clustering-based classification algorithm that possesses the advantages of both classification and clustering techniques for mixed data analysis. The proposed algorithm employs the K-prototypes algorithm with a genetic algorithm to optimize weights and centroids and utilizes the bagging method to build multiple classifiers, thereby enhancing classification performance. Furthermore, it incorporates four mutation mechanisms, including Gaussian, Cauchy, Levy, and single-point mutations, to explore optimal solutions. This study suggests using a 20% sampling ratio for the bootstrap sampling in the proposed algorithm. This ratio has been proven to be sufficient for achieving good classification performance while reducing computational time. Experimental results indicate that the proposed algorithm outperforms benchmark classifiers, demonstrating superior classification performance across five performance indicators. In addition, the loan eligibility case study offers valuable insights into applying the proposed algorithm in real-world scenarios, demonstrating that the proposed algorithm can achieve superior classification performance compared to other algorithms. It also offers managerial implications to help different industries and fields understand the appropriate timing and scenarios for implementing the algorithm.

Deep reinforcement learning based on balanced stratified prioritized experience replay for customer credit scoring in peer-to-peer lending

In recent years, deep reinforcement learning (DRL) models have been successfully utilised to solve various classification problems. However, these models have never been applied to customer credit scoring in peer-to-peer (P2P) lending. Moreover, the imbalanced class distribution in experience replay, which may affect the performance of DRL models, has rarely been considered. Therefore, this article proposes a novel DRL model, namely a deep Q-network based on a balanced stratified prioritized experience replay (DQN-BSPER) model, for customer credit scoring in P2P lending. Firstly, customer credit scoring is formulated as a discrete-time finite-Markov decision process. Subsequently, a balanced stratified prioritized experience replay technology is presented to optimize the loss function of the deep Q-network model. This technology can not only balance the numbers of minority and majority experience samples in the mini-batch by using stratified sampling technology but also select more important experience samples for replay based on the priority principle. To verify the model performance, four evaluation measures are introduced for the empirical analysis of two real-world customer credit scoring datasets in P2P lending. The experimental results show that the DQN-BSPER model can outperform four benchmark DRL models and seven traditional benchmark classification models. In addition, the DQN-BSPER model with a discount factor γ of 0.1 has excellent credit scoring performance.

Explicit implementation of hydrogen transport in metals

Hydrogen embrittlement prediction demands a numerical framework coupling a damage model with local hydrogen concentration. The inherent nonlinearity in hydrogen-stress-damage interactions challenges convergence in implicit schemes. To address this limitation, we propose a novel chemical potential-based explicit formulation for simulating hydrogen transport in metals. Our approach exploits a heat transfer analogy, linking mechanical and hydrogen transport via inelastic energy as a heat source. By employing chemical potential rather than lattice concentration, our method eliminates the need for user-defined boundary conditions and hydrostatic stress gradient determination. We integrate a VUMATHT subroutine for diffusion modelling and a VUMAT subroutine for material behaviour, coupling stress and strain rates as a heat source for diffusion. Validating against a classical benchmark, we compare our explicit approach with hydrogen concentration-based methods in ABAQUS Standard and Comsol Multiphysics. Stability conditions are assessed for different mesh sizes and mass scaling densities and the capabilities of our approach are showcased for 3D simulations of notched tensile specimens. Our framework offers a novel and efficient pathway for integrating hydrogen transport with user-defined material behaviour, promising advancements in hydrogen-informed damage models.

Symmetry-Enhanced, Improved Pathfinder Algorithm-Based Multi-Strategy Fusion for Engineering Optimization Problems

The pathfinder algorithm (PFA) starts with a random search for the initial population, which is then partitioned into only a pathfinder phase and a follower phase. This approach often results in issues like poor solution accuracy, slow convergence, and susceptibility to local optima in the PFA. To address these challenges, a multi-strategy fusion approach is proposed in the symmetry-enhanced, improved pathfinder algorithm-based multi-strategy fusion for engineering optimization problems (IPFA) for function optimization problems. First, the elite opposition-based learning mechanism is incorporated to improve the population diversity and population quality, to enhance the solution accuracy of the algorithm; second, to enhance the convergence speed of the algorithm, the escape energy factor is embedded into the prey-hunting phase of the GWO and replaces the follower phase in the PFA, which increases the diversity of the algorithm and improves the search efficiency of the algorithm; lastly, to solve the problem of easily falling into the local optimum, the optimal individual position is perturbed using the dimension-by-dimension mutation method of t-distribution, which helps the individual to jump out of the local optimum rapidly and advance toward other regions. The IPFA is used for testing on 16 classical benchmark test functions and 29 complex CEC2017 function sets. The final optimization results of PFA and IPFA in pressure vessels are 5984.8222 and 5948.3597, respectively. The final optimization results in tension springs are 0.012719 and 0.012699, respectively, which are comparable with the original algorithm and other algorithms. A comparison between the original algorithm and other algorithms shows that the IPFA algorithm is significantly enhanced in terms of solution accuracy, and the lower engineering cost further verifies the robustness of the IPFA algorithm.

Generalisation of a synchronous solution of the parity problem on cyclic configurations over a non-circulant graph

The parity problem is a classical binary benchmark for addressing the computational ability and limitations of automata networks. It refers to conceiving a local rule to allow deciding whether the number of 1-states in the nodes of an arbitrary network is an odd or even number, without global access to the nodes. In its standard formulation, the automata network has an odd number of nodes whose states, arranged as a cyclic configuration, should converge to a fixed point of all 0s, if the initial configuration has an even number of 1s, or to a fixed point of all 1s, otherwise. It has been shown that a local rule alone is able to solve the problem in this formulation, including a synchronous solution we have recently shown, totally based on the local parity rule of the cellular automata elementary space (number 150), with a certain pattern of connection between the nodes. Here, we generalise this solution, showing how to construct many others, combining rule 150 with rules 170 and 240, which are the local shifts of the same space, in such a way that the original solution is just one among countless possibilities. Such solutions may have different convergence times for specific configurations, but are equivalent in the context of all configurations of a given size; also, various solutions form symmetric pairs, in terms of the actual rules they used. The solutions were evaluated computationally and presented here without their formal proofs, but empirical evidences strongly suggest that they can be obtained by the same kind of technique we used in the solution exclusively with rule 150.

An Efficient Multi-Task Synergetic Network for Polyp Segmentation and Classification.

Colonoscopy is considered the best diagnostic tool for early detection and resection of polyps, which can effectively prevent consequential colorectal cancer. In clinical practice, segmenting and classifying polyps from colonoscopic images have a great significance since they provide precious information for diagnosis and treatment. In this study, we propose an efficient multi-task synergetic network (EMTS-Net) for concurrent polyp segmentation and classification, and we introduce a polyp classification benchmark for exploring the potential correlations of the above-mentioned two tasks. This framework is composed of an enhanced multi-scale network (EMS-Net) for coarse-grained polyp segmentation, an EMTS-Net (Class) for accurate polyp classification, and an EMTS-Net (Seg) for fine-grained polyp segmentation. Specifically, we first obtain coarse segmentation masks by using EMS-Net. Then, we concatenate these rough masks with colonoscopic images to assist EMTS-Net (Class) in locating and classifying polyps precisely. To further enhance the segmentation performance of polyps, we propose a random multi-scale (RMS) training strategy to eliminate the interference caused by redundant information. In addition, we design an offline dynamic class activation mapping (OFLD CAM) generated by the combined effect of EMTS-Net (Class) and RMS strategy, which optimizes bottlenecks between multi-task networks efficiently and elegantly and helps EMTS-Net (Seg) to perform more accurate polyp segmentation. We evaluate the proposed EMTS-Net on the polyp segmentation and classification benchmarks, and it achieves an average mDice of 0.864 in polyp segmentation and an average AUC of 0.913 with an average accuracy of 0.924 in polyp classification. Quantitative and qualitative evaluations on the polyp segmentation and classification benchmarks demonstrate that our EMTS-Net achieves the best performance and outperforms previous state-of-the-art methods in terms of both efficiency and generalization.

Automatic Analysis and Anomaly Detection System of Transverse Electron Beam Profile Based on Advanced and Interpretable Deep Learning Architectures

Abstract The National Synchrotron Radiation Center SOLARIS, ranked among the top infrastructures of that type worldwide, is the only one located in Central-Eastern Europe, in Poland. The SOLARIS Center, with six fully operational beamlines, serves as a hub for research across a diverse range of disciplines. This cutting-edge facility fosters innovation in fields like biology, chemistry, and physics as well as material engineering, nanotechnology, medicine, and pharmacology. With its advanced infrastructure and multidisciplinary approach, the SOLARIS Center enables discoveries and pushes the boundaries of knowledge. The most important aspect of such enormous research as well as industry infrastructure is to provide stable working conditions for the users and the conducted projects. Due to its unique properties, problem complexities, and challenges that require advanced approaches, the problem of anomaly detection and automatic analysis of signals for the beam stability assessment is still a huge challenge that has not been fully developed. To increase the effectiveness of centers with advanced research infrastructure we focus on the automatic analysis of transverse beam profiles generated by the Pinhole diagnostic beam-line. Pinhole beamlines are typically installed in the middle and high-energy synchrotrons to thoroughly analyze emitted X-rays and therefore assess electron beam quality. To address the problem we take advantage of AI solutions including up-to-date pre-trained convolutional neural network (CNN) models among others EfficientNetB0-B4-B6, InceptionV3 and DenseNet121. In this research, we propose the benchmark for Pinhole image classification including data preprocessing, model implementation, training process, hyperparameter selection as well as testing phase. Created from scratch database contains over one million transverse beam profile images. The proposed solution, based on the InceptionV3 architecture, classifies pinhole beamline images with 94.1% accuracy and 96.6% precision which is a state-of-the-art result in this research area. Finally, we employed interpretability algorithms to perform an analysis of the models and achieved results.

Robust visual tracking via modified Harris hawks optimization

Due to its outstanding efficiency and high precision, Harris hawks optimization (HHO for short) is suitable for solving the problem of visual target tracking under conditions of occlusion, deformation, rotation and in other complicated tracking scenes. A visual target tracker based on HHO is proposed in this study. To further promote the efficiency and stability of the standard HHO method and reduce the probability of the iteration falling into local optima and algorithm prematurity, in this study we propose an improved method called Super-HHO and apply it to visual target tracking. Compared with standard HHO, Super-HHO is superior due to its parameter optimization and updating strategy. We first optimize the random parameters of HHO via chaos theory to avoid frequent repeated exploration of the feasible region. Next, we design a nonlinear renewal strategy for the escape energy, which solves the problem in traditional HHO in which the fixed escape energy cannot accurately reflect the real hunting process of Harris hawks. Mutation strategies are also designed for the locations of the prey and the hunters to improve the optimization ability and eliminate the risk of falling into local extremes. In addition, a frame scale adjustment method model is developed to address the the issue in which the use of a size-fixed tracking frame makes it easy to include too many invalid features, which reduces the efficiency. Finally, we use the OTB2015, and VOT2018 tracking evaluation datasets, which contain hundreds of visual sequences and more than 10 complex interference scenes to conduct a qualitative analysis, a quantitative analysis and a statistical analysis of ours and other classic trackers, and to effectively test and compare the success ratio, precision and stability of each tracker. The proposed method was also compared with other classic trackers using classic large-scale benchmarks such as LaSOT and TrackingNet. Experimental data prove that ours performs well in terms of robustness, precision and efficiency.