Pruning Approaches Research Articles

Saliency and location aware pruning of deep visual detectors for autonomous driving

Despite the remarkable achievements of deep neural networks, their high computational complexity limits their wide use in many real-world embedded applications, such as autonomous driving perception. While current neural network pruning approaches can reduce model complexity to various extents, they often adopt local or ad hoc importance measures that are not directly related to the final task. More importantly, most of them focus on classification tasks and do not take location information into consideration during pruning. To address these issues, we present a novel channel importance measure that incorporates detection-related saliency and location awareness, specifically designed for the pruning of self-driving visual detectors. Our comprehensive experiments on the KITTI and COCO_traffic datasets demonstrate that our pruning method achieves significant reductions in model size and computational operations with little performance degradation. Notably, it outperforms other state-of-the-art methods across various pruning rates and base detectors. Our pruned YOLOX-S model with 40.2% fewer parameters even improves the original model’s mAP by 1.8% on KITTI. Moreover, we experimentally highlight the potential of our pruning approach in effectively detecting small-scale objects.

A class-incremental learning approach for learning feature-compatible embeddings

Humans have the ability to constantly learn new knowledge. However, for artificial intelligence, trying to continuously learn new knowledge usually results in catastrophic forgetting, the existing regularization-based and dynamic structure-based approaches have shown great potential for alleviating. Nevertheless, these approaches have certain limitations. They usually do not fully consider the problem of incompatible feature embeddings. Instead, they tend to focus only on the features of new or previous classes and fail to comprehensively consider the entire model. Therefore, we propose a two-stage learning paradigm to solve feature embedding incompatibility problems. Specifically, we retain the previous model and freeze all its parameters in the first stage while dynamically expanding a new module to alleviate feature embedding incompatibility questions. In the second stage, a fusion knowledge distillation approach is used to compress the redundant feature dimensions. Moreover, we propose weight pruning and consolidation approaches to improve the efficiency of the model. Our experimental results obtained on the CIFAR-100, ImageNet-100 and ImageNet-1000 benchmark datasets show that the proposed approaches achieve the best performance among all the compared approaches. For example, on the ImageNet-100 dataset, the maximal accuracy improvement is 5.08%. Code is available at https://github.com/ybyangjing/CIL-FCE.

CATRO: Channel Pruning via Class-Aware Trace Ratio Optimization.

Deep convolutional neural networks are shown to be overkill with high parametric and computational redundancy in many application scenarios, and an increasing number of works have explored model pruning to obtain lightweight and efficient networks. However, most existing pruning approaches are driven by empirical heuristics and rarely consider the joint impact of channels, leading to unguaranteed and suboptimal performance. In this article, we propose a novel channel pruning method via c lass-aware t race r atio o ptimization (CATRO) to reduce the computational burden and accelerate the model inference. Utilizing class information from a few samples, CATRO measures the joint impact of multiple channels by feature space discriminations and consolidates the layerwise impact of preserved channels. By formulating channel pruning as a submodular set function maximization problem, CATRO solves it efficiently via a two-stage greedy iterative optimization procedure. More importantly, we present theoretical justifications on convergence of CATRO and performance of pruned networks. Experimental results demonstrate that CATRO achieves higher accuracy with similar computation cost or lower computation cost with similar accuracy than other state-of-the-art channel pruning algorithms. In addition, because of its class-aware property, CATRO is suitable to prune efficient networks adaptively for various classification subtasks, enhancing handy deployment and usage of deep networks in real-world applications.

A roulette wheel-based pruning method to simplify cumbersome deep neural networks

Deep neural networks (DNNs) have been applied in many pattern recognition or object detection applications. DNNs generally consist of millions or even billions of parameters. These demanding computational storage and requirements impede deployments of DNNs in resource-limited devices, such as mobile devices, micro-controllers. Simplification techniques such as pruning have commonly been used to slim DNN sizes. Pruning approaches generally quantify the importance of each component such as network weight. Weight values or weight gradients in training are commonly used as the importance metric. Small weights are pruned and large weights are kept. However, small weights are possible to be connected with significant weights which have impact to DNN outputs. DNN accuracy can be degraded significantly after the pruning process. This paper proposes a roulette wheel-like pruning algorithm, in order to simplify a trained DNN while keeping the DNN accuracy. The proposed algorithm generates a branch of pruned DNNs which are generated by a roulette wheel operator. Similar to the roulette wheel selection in genetic algorithms, small weights are more likely to be pruned but they can be kept; large weights are more likely to be kept but they can be pruned. The slimmest DNN with the best accuracy is selected from the branch. The performance of the proposed pruning algorithm is evaluated by two deterministic datasets and four non-deterministic datasets. Experimental results show that the proposed pruning algorithm generates simpler DNNs while DNN accuracy can be kept, compared to several existing pruning approaches.

Fluctuation-Based Adaptive Structured Pruning for Large Language Models

Network Pruning is a promising way to address the huge computing resource demands of the deployment and inference of Large Language Models (LLMs). Retraining-free is important for LLMs' pruning methods. However, almost all of the existing retraining-free pruning approaches for LLMs focus on unstructured pruning, which requires specific hardware support for acceleration. In this paper, we propose a novel retraining-free structured pruning framework for LLMs, named FLAP (FLuctuation-based Adaptive Structured Pruning). It is hardware-friendly by effectively reducing storage and enhancing inference speed. For effective structured pruning of LLMs, we highlight three critical elements that demand the utmost attention: formulating structured importance metrics, adaptively searching the global compressed model, and implementing compensation mechanisms to mitigate performance loss. First, FLAP determines whether the output feature map is easily recoverable when a column of weight is removed, based on the fluctuation pruning metric. Then it standardizes the importance scores to adaptively determine the global compressed model structure. At last, FLAP adds additional bias terms to recover the output feature maps using the baseline values. We thoroughly evaluate our approach on a variety of language benchmarks. Without any retraining, our method significantly outperforms the state-of-the-art methods, including LLM-Pruner and the extension of Wanda in structured pruning. The code is released at https://github.com/CASIA-IVA-Lab/FLAP.

Domain Adaptive Channel Pruning

Domain adaptation is an effective approach to improve the generalization ability of deep learning methods, which makes a deep model more stable and robust. However, these methods often suffer from a deployment problem when deep models are deployed on different types of edge devices. In this work, we propose a new channel pruning method called Domain Adaptive Channel Pruning (DACP), which is specifically designed for the unsupervised domain adaptation task, where there is considerable data distribution mismatch between the source and the target domains. We prune the channels and adjust the weights in a layer-by-layer fashion. In contrast to the existing layer-by-layer channel pruning approaches that only consider how to reconstruct the features from the next layer, our approach aims to minimize both classification error and domain distribution mismatch. Furthermore, we propose a simple but effective approach to utilize the unlabeled data in the target domain. Our comprehensive experiments on two benchmark datasets demonstrate that our newly proposed DACP method outperforms the existing channel pruning approaches under the unsupervised domain adaptation setting.

Hardware-Aware Evolutionary Explainable Filter Pruning for Convolutional Neural Networks

Filter pruning of convolutional neural networks (CNNs) is a common technique to effectively reduce the memory footprint, the number of arithmetic operations, and, consequently, inference time. Recent pruning approaches also consider the targeted device (i.e., graphics processing units) for CNN deployment to reduce the actual inference time. However, simple metrics, such as the ℓ1\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\ell ^1$$\\end{document}-norm, are used for deciding which filters to prune. In this work, we propose a hardware-aware technique to explore the vast multi-objective design space of possible filter pruning configurations. Our approach incorporates not only the targeted device but also techniques from explainable artificial intelligence for ranking and deciding which filters to prune. For each layer, the number of filters to be pruned is optimized with the objective of minimizing the inference time and the error rate of the CNN. Experimental results show that our approach can speed up inference time by 1.40× and 1.30× for VGG-16 on the CIFAR-10 dataset and ResNet-18 on the ILSVRC-2012 dataset, respectively, compared to the state-of-the-art ABCPruner.

An analysis of ensemble pruning methods under the explanation of Random Forest

“Black box” models created by modern machine learning techniques are typically hard to interpret. Thus, the necessity of explainable artificial intelligence (XAI) has grown for understanding the rationale behind those models and converting them into white boxes. Random Forest is a black box model essential in various domains due to its flexibility, ease of use, and remarkable predictive performance. One method for explaining a Random Forest is transforming it into a self-explainable Decision Tree using Forest-Based Tree (FBT) algorithm. It basically consists of three main phases, pruning, conjunction set generation, and Decision Tree construction. In this paper, we examine six state-of-the-art pruning approaches and analyze their effect on FBT performance through pruned FBT (PFBT) in order to minimize its computational complexity. This would make it appropriate for forests and datasets of any size. They are assessed on 30 datasets, and the results show that UMEP and Hybrid pruning methods can be effectively used in the pruning stage of the PFBT algorithm in terms of pruning time and predictive performance. However, the AUC-Greedy method achieves good performance with small-size datasets.

Extremely Sparse Networks via Binary Augmented Pruning for Fast Image Classification.

Network pruning and binarization have been demonstrated to be effective in neural network accelerator design for high speed and energy efficiency. However, most existing pruning approaches achieve a poor tradeoff between accuracy and efficiency, which on the other hand, has limited the progress of neural network accelerators. At the same time, binary networks are highly efficient, however, a large accuracy gap exists between binary networks and their full-precision counterparts. In this article, we investigate the merits of extremely sparse networks with binary connections for image classification through software-hardware codesign. More specifically, we first propose a binary augmented extremely pruning method that can achieve ~98% sparsity with small accuracy degradation. Then we design the hardware architecture based on the resulting sparse and binary networks, which extensively explores the benefits of extreme sparsity with negligible resource consumption introduced by binary branch. Experiments on large-scale ImageNet classification and field-programmable gate array (FPGA) demonstrate that the proposed software-hardware architecture can achieve a prominent tradeoff between accuracy and efficiency.

Late Pruning and Forced Vine Regrowth in Chardonnay and Pinot Noir: Benefits and Drawbacks in the Trento DOC Basin (Italy)

Targeted pruning techniques may have the potential to shift phenological progression in grapevines. These approaches can be useful to escape late frost (after budburst) and to delay ripening, maintaining yield and berry quality under specific environmental scenarios and for targeted oenological aims. In this work we tested a series of pruning methods (late pruning and different forcing approaches to unlock the para-dormant buds) with the primary objective of delaying phenological onset while increasing must quality components in the Trento DOC basin (Italy) over three field seasons either in cv Chardonnay or cv Pinot noir. Delayed pruning shifted the onset of most of the phenological stages by 3 to 10 days with a general increase in must acidity at harvest over two seasons. Forcing techniques, and in particular forcing vine regrowth (i.e., removing primary productivity along with leaves and secondary shoots to force dormant bud growth), substantially reduced yield per vine (p < 0.001) but significantly (p < 0.001) enhanced total acidity and yeast available nitrogen when compared to winter pruning vines in both Chardonnay and Pinot noir. Different spring pruning for unlocking para-dormant buds aimed at double-cropping in 2022 did not yield a secondary harvest potentially due to lack of vigor in the vineyard (Pinot noir) and unfavorable environmental conditions that putatively anticipated bud pre-dormancy. However, reduced sugar levels in the primary productivity bunches of the treated vines were observed, suggesting that early source limitation (e.g., at fruit set) applied via reduced active photosynthetic leaf area may slow down technological ripening even if compensatory behaviors were observed in the treated plants through enhanced stomatal conductance. Our work provides evidence of the potential usefulness of different pruning approaches for manipulating phenological progression and berry ripening dynamics under climate change scenario in the Trento DOC area and suggests a lack of efficacy and specific drawbacks particularly in dry and hot years.

Pruning Coherent Integrated Photonic Neural Networks

Coherent integrated photonic neural networks (IPNNs) are increasingly being explored for rapidly growing artificial intelligence applications. However, the principal roadblocks in the scalability of IPNNs are their large area footprint and high tuning power consumption during both training and inferencing. In deep neural networks (DNNs), software techniques to prune redundant weights are often utilized to reduce resource (e.g., memory, computation, and power) overheads. However, due to the complex nature in which the software weights are mapped to the building blocks of IPNNs, prior efforts to apply existing pruning approaches to IPNNs have been ineffective. We present CHAMP and LTPrune, two novel hardware-aware pruning techniques for IPNNs. Using a case study of three IPNNs with different footprints, we show that both these methods can prune more than 99% of the phase angles (which are similar to the weight parameters in DNNs). We also analyze the performance of the pruned IPNNs under phase uncertainties and present a comparative analysis of the two methods to enable advanced hardware-software-assisted design-optimization techniques for IPNNs. To expedite pruning, we also propose HybridPrune, where CHAMP and LTPrune are used in conjunction to obtain similar network sparsity as standalone-LTPrune but with up to 78.3% fewer retraining epochs.

Compression of Deep Convolutional Neural Network Using Additional Importance-Weight-Based Filter Pruning Approach

The success of the convolutional neural network (CNN) comes with a tremendous growth of diverse CNN structures, making it hard to deploy on limited-resource platforms. These over-sized models contain a large amount of filters in the convolutional layers, which are responsible for almost 99% of the computation. The key question here arises: Do we really need all those filters? By removing entire filters, the computational cost can be significantly reduced. Hence, in this article, a filter pruning method, a process of discarding a subset of unimportant or weak filters from the original CNN model, is proposed, which alleviates the shortcomings of over-sized CNN architectures at the cost of storage space and time. The proposed filter pruning strategy is adopted to compress the model by assigning additional importance weights to convolutional filters. These additional importance weights help each filter learn its responsibility and contribute more efficiently. We adopted different initialization strategies to learn more about filters from different aspects and prune accordingly. Furthermore, unlike existing pruning approaches, the proposed method uses a predefined error tolerance level instead of the pruning rate. Extensive experiments on two widely used image segmentation datasets: Inria and AIRS, and two widely known CNN models for segmentation: TernausNet and standard U-Net, verify that our pruning approach can efficiently compress CNN models with almost negligible or no loss of accuracy. For instance, our approach could significantly reduce 85% of all floating point operations (FLOPs) from TernausNet on Inria with a negligible drop of 0.32% in validation accuracy. This compressed network is six-times smaller and almost seven-times faster (on a cluster of GPUs) than that of the original TernausNet, while the drop in the accuracy is less than 1%. Moreover, we reduced the FLOPs by 84.34% without significantly deteriorating the output performance on the AIRS dataset for TernausNet. The proposed pruning method effectively reduced the number of FLOPs and parameters of the CNN model, while almost retaining the original accuracy. The compact model can be deployed on any embedded device without any specialized hardware. We show that the performance of the pruned CNN model is very similar to that of the original unpruned CNN model. We also report numerous ablation studies to validate our approach.

Channel pruning based on convolutional neural network sensitivity

Pruning is a useful technique for decreasing the memory consumption and floating point operations (FLOPs) of deep convolutional neural network (CNN) models. Nevertheless, at modest pruning levels, current structured pruning approaches often lead to considerable declines in accuracy. Furthermore, existing approaches often treat pruning rates as super parameters, neglecting the sensitivity of different convolution layers. In this study, we propose a novel sensitivity-based method for channel pruning that utilizes second-order sensitivity as a criterion. The essential concept is to prune insensitive filters while retaining sensitive ones. We quantify the sensitivity of the filter using the sum of the sensitivities of all weights in the filter, rather than the magnitude-based metric frequently applied in the literature. Furthermore, a layer sensitivity approach based on the Hessian eigenvalues of each layer is introduced into the process of automatically choosing the most appropriate pruning rate for each layer. Experiments on a variety of modern CNN architectures demonstrate that we can considerably enhance the pruning rate while sacrificing a small amount of accuracy, resulting in a reduction of more than 60% in FLOPs on CIFAR-10. Notably, on ImageNet, pruning based on ResNet50 decreased the FLOPs by 56.3% while losing only 0.92% of accuracy.

Design Space Exploration of a Sparse MobileNetV2 Using High-Level Synthesis and Sparse Matrix Techniques on FPGAs.

Convolution Neural Networks (CNNs) are gaining ground in deep learning and Artificial Intelligence (AI) domains, and they can benefit from rapid prototyping in order to produce efficient and low-power hardware designs. The inference process of a Deep Neural Network (DNN) is considered a computationally intensive process that requires hardware accelerators to operate in real-world scenarios due to the low latency requirements of real-time applications. As a result, High-Level Synthesis (HLS) tools are gaining popularity since they provide attractive ways to reduce design time complexity directly in register transfer level (RTL). In this paper, we implement a MobileNetV2 model using a state-of-the-art HLS tool in order to conduct a design space exploration and to provide insights on complex hardware designs which are tailored for DNN inference. Our goal is to combine design methodologies with sparsification techniques to produce hardware accelerators that achieve comparable error metrics within the same order of magnitude with the corresponding state-of-the-art systems while also significantly reducing the inference latency and resource utilization. Toward this end, we apply sparse matrix techniques on a MobileNetV2 model for efficient data representation, and we evaluate our designs in two different weight pruning approaches. Experimental results are evaluated with respect to the CIFAR-10 data set using several different design methodologies in order to fully explore their effects on the performance of the model under examination.

M-CORE: A Novel Approach for Land Cover Fraction Mapping Using Multisite Spectral Libraries

The concept of a generic spectral library (GSL), i.e., a structured collection of image spectra sampled from various optical remote sensing images covering different sites and points in time, has been suggested in previous studies to facilitate the tedious process of producing training data for remote sensing-based land cover (LC) mapping. When using multisource libraries collected over different sites, library pruning approaches are needed to extract an apt set of labeled spectra to perform mapping on a specific area. Library pruning mainly aims to discard image-irrelevant and redundant spectra, while limiting spectral confusion during the mapping process. Most library pruning approaches focus only on one of these aspects, which has been shown to negatively affect mapping accuracies. This article emphasizes the need for a multistep approach to optimize a GSL for site-specific mapping. We propose a new library pruning method called M-CORE, specifically designed to facilitate LC fraction mapping. The method extends multiple signal classification (MUSIC) with a confusion reduction (CORE) component. Vegetation-impervious-soil (VIS) fraction mapping experiments on a Sentinel-2 image of Brussels, using a GSL with and without local spectra, show the added value of M-CORE over previously proposed library pruning methods and demonstrate the feasibility of GSL-based mapping in an urban context.

Small Network for Lightweight Task in Computer Vision: A Pruning Method Based on Feature Representation.

Many current convolutional neural networks are hard to meet the practical application requirement because of the enormous network parameters. For accelerating the inference speed of networks, more and more attention has been paid to network compression. Network pruning is one of the most efficient and simplest ways to compress and speed up the networks. In this paper, a pruning algorithm for the lightweight task is proposed, and a pruning strategy based on feature representation is investigated. Different from other pruning approaches, the proposed strategy is guided by the practical task and eliminates the irrelevant filters in the network. After pruning, the network is compacted to a smaller size and is easy to recover accuracy with fine-tuning. The performance of the proposed pruning algorithm is validated on the acknowledged image datasets, and the experimental results prove that the proposed algorithm is more suitable to prune the irrelevant filters for the fine-tuning dataset.

Memory-Free Stochastic Weight Averaging by One-Way Variational Pruning

Recent works on convolutional neural networks (CNN) have attempted to find the local optima with ensemble-based approaches. Fast Geometric Ensemble (FGE) showed that captured weight points at the end of training time circulate local optima. This led to the Stochastic Weight Averaging (SWA) approach, which averages multiple model weights to find the local optima. However, they are limited by their output of fully-parameterized models, including needless parameters, after the training procedure. To solve this problem, we propose a novel training procedure: Stochastic Weight Averaging by One-way Variational Pruning (SWA-OVP). SWA-OVP reduces the number of model parameters by variationally updating the mask of weights for pruning. SWA-OVP variationally generates a mask for pruned weights in each iteration while recent pruning approaches produce the mask at the end of each training. In addition, our SWA-OVP prunes the model in a one-way training procedure, while other recent approaches prune the model weights in iterative training or require additional computation. Our experiment shows that SWA-OVP using only a 0.5x% $\sim$ 0.7x% parameter size achieves even higher accuracy than SWA and FGE on several networks, such as Pre-ResNet110, Pre-ResNet164 and WideResNet28x10 on CIFAR10 and CIFAR100 datasets. SWA-OVP also achieves better performance compared to state-of-the-art pruning approaches.

Task dependent deep LDA pruning of neural networks

With deep learning’s success, a limited number of popular deep nets have been widely adopted for various vision tasks. However, this usually results in unnecessarily high complexities and possibly many features of low task utility. In this paper, we address this problem by introducing a task-dependent deep pruning framework based on Fisher’s Linear Discriminant Analysis (LDA). The approach can be applied to convolutional, fully-connected, and module-based deep network structures, in all cases leveraging the high decorrelation of neuron motifs found in the pre-decision space and cross-layer deconv dependency. Moreover, we examine our approach’s potential in network architecture search for specific tasks and analyze the influence of our pruning on model robustness to noises and adversarial attacks. Experimental results on datasets of generic objects (ImageNet, CIFAR100) as well as domain specific tasks (Adience, and LFWA) illustrate our framework’s superior performance over state-of-the-art pruning approaches and fixed compact nets (e.g. SqueezeNet, MobileNet). The proposed method successfully maintains comparable accuracies even after discarding most parameters (98%–99% for VGG16, up to 82% for the already compact InceptionNet) and with significant FLOP reductions (83% for VGG16, up to 64% for InceptionNet). Through pruning, we can also derive smaller, but more accurate and more robust models suitable for the task.

Jojoba pruning: New practices to rejuvenate the plant, improve yield and reduce alternate bearing

Commercial pruning practice in jojoba plantations is traditionally dictated by maintenance requirements only. It enables machine movement between the plants and efficient harvest but is not designed to maximize long-term productivity. In this study, mechanical and manual pruning approaches were tested in two cultivars in a mature jojoba plantation, in comparison to the common practice. These new approaches were designed to enable better light penetration into the canopy, aiming to improve growth and productivity. As jojoba is an alternate bearing crop, the vegetative and reproductive performances were observed over four years, by remote sensing and manual measurements. The pruning type and strategy had a significant effect on growth, with a distinction between the two tested cultivars. Top-pruning methods were found to best encourage new branching and yield, while side-pruning practices were less effective. Several treatments, including hedge pruning, attenuated or eliminated the alternation cycle. We conclude that using the proposed pruning practices would be beneficial in jojoba cultivation, and that the specific method should be suited to the cultivar characteristics.

Multistep Prediction-Based Adaptive Dynamic Programming Sensor Scheduling Approach for Collaborative Target Tracking in Energy Harvesting Wireless Sensor Networks

Sensor scheduling for energy-efficient collaborative target tracking in wireless sensor networks (WSNs) is an important problem to deal with the limited network resources. With the recent development and emerging applications of energy acquisition technologies, it has become possible to overcome the bottleneck of battery energy in WSNs using the energy harvesting devices, where theoretically the lifetime of the network could be extended to the infinite. However, the energy harvesting WSN also poses new challenges for sensor scheduling algorithm over the infinite horizon under the limited sensor energy harvesting capabilities. In this article, a novel multistep prediction-based adaptive dynamic programming (MSPADP) approach is proposed for collaborative target tracking in energy harvesting WSNs to schedule sensors over an infinite horizon, according to the ADP mechanism. The “action” module of MSPADP is designed to obtain the sensor scheduling for multiple steps starting from the current step, and implemented by the minimal-cost first search (MCFS) decision tree scheme, and the “critic network” module of MSPADP is iteratively performed to optimize the performance for the remaining infinite steps using neural network. Extended Kalman filter (EKF) is adopted to predict and estimate the target state. The performance index is defined by the tracking accuracy derived from EKF and the energy consumption predicted by the candidate sensor schedule. Theoretical analysis shows the optimality of MSPADP, and simulation results demonstrate its superior tracking performance compared with single-step prediction-based ADP (SSPADP), multistep prediction-based dynamic programming (MSPDP), and multistep prediction-based pruning (MSPP) sensor scheduling approaches. Note to Practitioners-Collaborative target tracking is a typical problem in wireless sensor networks (WSNs) where the sensors need to be scheduled to address the constraints of the limited network resources, such as sensor energy usually supplied by the battery. In the recent years, energy harvesting device has been developed and applied to WSNs to overcome the energy restriction. As the energy harvesting capabilities of the sensors are limited, sensor scheduling remains as a challenging problem and is studied in this article. A novel multistep prediction-based adaptive dynamic programming (MSPADP) approach is proposed for collaborative target tracking, by scheduling sensors for the current time step based on the predictions of the subsequent steps over an infinite horizon. It runs iteratively in two modules: obtaining the previous optimal multistep sensor scheduling and updating the remaining infinite-step performance. Simulation results show its superior tracking performance compared with single-step prediction-based ADP (SSPADP), multistep prediction-based dynamic programming (MSPDP), and multistep prediction-based pruning (MSPP) approaches, and lay a good foundation for the practical applications.