Low Space Complexity Research Articles

Improved multi-focus image fusion using online convolutional sparse coding based on sample-dependent dictionary

Multi-focus image fusion merges multiple images captured from different focused regions of a scene to create a fully-focused image. Convolutional sparse coding (CSC) methods are commonly employed for accurate extraction of focused regions, but they often disregard computational costs. To overcome this, an online convolutional sparse coding (OCSC) technique was introduced, but its performance is still limited by the number of filters used, affecting overall performance negatively. To address these limitations, a novel approach called Sample-Dependent Dictionary-based Online Convolutional Sparse Coding (SCSC) was proposed. SCSC enables the utilization of additional filters while maintaining low time and space complexity for processing high-dimensional or large data. Leveraging the computational efficiency and effective global feature extraction of SCSC, we propose a novel method for multi-focus image fusion. Our method involves a two-layer decomposition of each source image, yielding a base layer capturing the predominant features and a detail layer containing finer details. The amalgamation of the fused base and detail layers culminates in the reconstruction of the final image. The proposed method significantly mitigates artifacts, preserves fine details at the focus boundary, and demonstrates notable enhancements in both visual quality and objective evaluation of multi-focus image fusion.

Fast linear model trees by PILOT

Linear model trees are regression trees that incorporate linear models in the leaf nodes. This preserves the intuitive interpretation of decision trees and at the same time enables them to better capture linear relationships, which is hard for standard decision trees. But most existing methods for fitting linear model trees are time consuming and therefore not scalable to large data sets. In addition, they are more prone to overfitting and extrapolation issues than standard regression trees. In this paper we introduce PILOT, a new algorithm for linear model trees that is fast, regularized, stable and interpretable. PILOT trains in a greedy fashion like classic regression trees, but incorporates an L2 boosting approach and a model selection rule for fitting linear models in the nodes. The abbreviation PILOT stands for PIecewise Linear Organic Tree, where ‘organic’ refers to the fact that no pruning is carried out. PILOT has the same low time and space complexity as CART without its pruning. An empirical study indicates that PILOT tends to outperform standard decision trees and other linear model trees on a variety of data sets. Moreover, we prove its consistency in an additive model setting under weak assumptions. When the data is generated by a linear model, the convergence rate is polynomial.

Kurdish Sign Language Recognition Using Pre-Trained Deep Learning Models

In the tapestry of rich human communication, sign language gleams like one of the basic threads of this art, giving voice to hundreds of deaf and hard-of-hearing individuals in the region. The technology for recognizing and translating sign language fell far behind what these communities needed. Therefore, the present investigation is going to compare the performance of three top-performing deep learning algorithms in recognizing the signs created from the database of Kurdish Sign Language. The models are going to be put to a rigorous test, with a variety of signs drawn. All the 3 models give a performance that is—from all indications—good or even excellent, this is MobileNetV2, a very good candidate that manages to walk an amazing line between the requirements of high accuracy, low space complexity, and acceptable time complexity. We conclude by looking at some exciting opportunities for future research, including integrating our models into hardware devices and expanding our study to a larger variety of sign languages. And just as any good journey would, it throws up as many questions as it answers, leaving us inspired by the many possibilities that will need to be explored to enhance communication for all.

Algorithm 1038: KCC: A MATLAB Package for k -Means-based Consensus Clustering

Consensus clustering is gaining increasing attention for its high quality and robustness. In particular, k -means-based Consensus Clustering (KCC) converts the usual computationally expensive problem to a classic k -means clustering with generalized utility functions, bringing potentials for large-scale data clustering on different types of data. Despite KCC’s applicability and generalizability, implementing this method such as representing the binary dataset in the k -means heuristic is challenging and has seldom been discussed in prior work. To fill this gap, we present a MATLAB package, KCC, that completely implements the KCC framework and utilizes a sparse representation technique to achieve a low space complexity. Compared to alternative consensus clustering packages, the KCC package is of high flexibility, efficiency, and effectiveness. Extensive numerical experiments are also included to show its usability on real-world datasets.

Ensemble model for accuracy prediction of protein secondary structure

<span>Predicting a protein’s secondary structure is crucial for understanding the working of proteins. Despite advancements over the years, the top predictors have achieved only 80% Q8 accuracy when sequence profile information is the sole input. An ensemble approach is proposed using convolutional neural network (CNN) and a classifier known as support vector machine (SVM) on both the partial and the whole CullPDB datasets. The protein secondary structure (PSS) has a complex hierarchical structure, as well as the ability to take into account the reliance between neighbouring labels. A detailed experiment yielding high levels of Q8 accuracy with scores of 97.91%, 85.13%, and 78.02% using 20%, 80%, and 100% respectively of the protein residues on the new predicted dataset CullPDB6133 which is better than the accuracies predicted by similar models. The proposed methodology highlights the use of CNN as a general framework, for efficiently predicting eight-state (Q8) accuracy of secondary protein structures with a low time and space complexity.</span>

The Probabilistic Image Encryption Algorithm Based on Galois Field GF(257)

Image encryption technology is an essential measure to protect image information from illegal access. A new probabilistic image encryption algorithm with 480-bit external secret key was proposed in this paper. A chaotic pseudo-random generator is employed to produce the equivalent keys. Then the plain image is covered by the equivalent keys under exclusive or operation to obtain a pending sequence. The pending sequence is transformed into a cipher sequence by arithmetic operations based on GF(257). Finally, the cipher sequence is also covered by the equivalent keys to obtain the cipher image. In the transformation, a 1024-bit pseudo-random number sequence is introduced as the base numbers for the power operation to assist the encryption. The proposed image encryption algorithm has the characteristic of probabilistic encryption. Encrypting the same plain image with the same secret key will result in completely different cipher images. Experiments show that the proposed algorithm has the low time complexity (9.67pi/bit) and space complexity (13.07bit/bit), and high sensitivities on the auxiliary sequence (49.94%), secret key (BACI 26.7698% typically), and plain image (BACI 26.7845% typically), and is a preferred image encryption scheme for commercial and personal image information security.

Privacy-preserving DeepFake face image detection

All the existing models for DeepFake detection focus on plaintext faces. However, outsourced computing is usually considered in practical applications for DeepFake detection and the input data may contain private and sensitive information. Thus, a privacy-preserving model named Secure DeepFake Detection Network (SecDFDNet) is proposed for the first time in this paper. The SecDFDNet uses the additive secret sharing method for secure DeepFake face detection. Specifically, firstly, some multi-party secure interaction protocols are designed for non-linear activation functions, i.e., SecReLU for ReLU function, SecSigm for sigmoid function, SecSpatial for spatial attention, and SecChannel for channel attention. Their security is proved in theory. Our protocols have low communication and space complexity. Then, the SecDFDNet model is proposed by using the designed secure protocols and trained plaintext DeepFake detection network (DFDNet). The experimental results show that the proposed SecDFDNet can detect DeepFake faces without revealing anything of private input, achieve the same accuracies as the plaintext DFDNet and outperform some existing models. The source code is available at https://github.com/imagecbj/Privacy-Preserving-DeepFake-Face-Image-Detection.

Transformer in optronic neural networks for image classification

Transformer, with its self-attention mechanism, can acquire global features while parallelizing training. This study proposes an Optronic Vision Transformer (OPViT), which utilizes spatial light modulators (SLMs) and lens group to realize matrix multiplication. This is the first time that Transformer has been introduced into optical neural network, fully exploiting the advantages of photonic computing to minimize its computational cost. Based on the advantages of OPViT, it is reconfigurable, scalable, and has low space complexity. We achieve excellent classification results directly with only one layer of Transformer and two or three layers of optical convolution, achieving 98.70% and 88.93% test accuracy on the MNIST and Fashion-MNIST datasets, respectively. The classification results outperform existing optical convolutional neural networks, and are even comparable to previous optical architectures connected to electronic networks.

A lightweight super-resolution network with skip-connections.

In some hospitals in remote areas, due to the lack of MRI scanners with high magnetic field intensity, only low-resolution MRI images can be obtained, hindering doctors from making correct diagnoses. In our study, higher-resolution images were obtained through low-resolution MRI images. Moreover, as our algorithm is a lightweight algorithm with a small number of parameters, it can be carried out in remote areas under the condition of the lack of computing resources. Moreover, our algorithm is of great clinical significance in providing references for doctors' diagnoses and treatment in remote areas. We compared different super-resolution algorithms to obtain high-resolution MRI images, including SRGAN, SPSR, and LESRCNN. A global skip connection was applied to the original network of LESRCNN to use global semantic information to get better performance. Experiments reported that our network improved SSMI by 0.8% and also achieved an obvious increase in PSNR, PI, and LPIPS compared to LESRCNN in our dataset. Similar to LESRCNN, our network has a very short running time, a small number of parameters, low time complexity, and low space complexity while ensuring high performance compared to SRGAN and SPSR. Five MRI doctors were invited for a subjective evaluation of our algorithm. All agreed on significant improvements and that our algorithm could be used clinically in remote areas and has great value. The experimental results demonstrated the performance of our algorithm in super-resolution MRI image reconstruction. It allows us to obtain high-resolution images in the absence of high-field intensity MRI scanners, which have great clinical significance. The short running time, a small number of parameters, low time complexity, and low space complexity ensure that our network can be used in grassroots hospitals in remote areas that lack computing resources. We can reconstruct high-resolution MRI images in a short time, thus saving time for patients. Our algorithm can be biased towards practical applications; however, doctors have affirmed the clinical value of our algorithm.

Towards time-constrained task allocation in semi-opportunistic mobile crowdsensing

Mobile crowdsensing (MCS) is an emerging sensing paradigm that leverages mobile users carrying smart devices to perform sensing tasks. On the one hand, as the sensing scenarios become more realistic and complex, the impact of time attributes on task allocation is significantly increased. However, most of the existing works only treat time attributes as supplementary constraints and ignore the existence of multiple time constraints. On the other hand, considering the exploration and application of emerging models, making MCS more adaptive and diversified also becomes a research focus. In this paper, we propose the Multiple Time constrained Task Allocation problem in Semi-Opportunistic mobile crowdsensing (SO-MTTA), with the goal of maximizing the sensing value obtained by the platform. We prove that the SO-MTTA problem is NP-hard and design a Multi-Layer Genetic Algorithm (MLGA), which uses a low space complexity encoding method for three layers: User-path-task, and adds a conflict elimination operation to correct possible conflicts of constraints. Finally, we conduct experiments on both synthetic and real-world datasets to successfully verify the effectiveness of our proposed algorithm.

A quantum algorithm for solving weapon target assignment problem

Quantum computers, known to have the potential for exponential speedup in solving some problems due to their superposition property, are expected to facilitate the solution of NP-hard optimisation problems. This study proposes a quantum algorithm to solve the weapon target assignment problem (WTAP), one of the NP-hard optimisation problems. The proposed quantum algorithm is a gate-based approach, and scenario examples are executed on Qiskit quantum computing platform and IBM Lima quantum computer with Falcon r4T processor type. The results manifest that the proposed quantum algorithm has low space and time complexity, demonstrating its memory and computational resources efficiency.

Fast Image Classification Algorithms Based on Sequential Analysis

This work is devoted to fast image recognition techniques based on statistical sequential analysis. We examine the possibility to sequentially process principal components and organize a convolutional neural network with early exits. We pay special attention to sequentially learning a multi-task lightweight neural network model in order to predict several facial attributes (age, gender, and ethnicity) based on preliminary training on the face classification task. We specifically note that the entire above-mentioned model should be fine-tuned in order to deal with the emotion recognition problem. Experimental evaluation on several datasets demonstrates that the proposed approach has high accuracy and very low running time and space complexity compared to known state-of-the-art methods.

Factorized multi-Graph matching

In recent years, multi-graph matching has become a popular yet challenging task in graph theory. There exist two major problems in multi-graph matching, i.e., the cycle-consistency problem, and the high time and space complexity problem. On one hand, the pairwise-based multi-graph matching methods are of low time and space complexity, but in order to keep the cycle-consistency of the matching results, they need additional constraints. Besides, the accuracy of the pairwise-based multi-graph matching is highly dependent on the selected optimization algorithms. On the other hand, the tensor-based multi-graph matching methods can avoid the cycle-consistency problem, while their time and space complexity is extremely high. In this paper, we found the equivalence between the pairwise-based and the tensor-based multi-graph matching methods under some specific circumstances. Based on this finding, we proposed a new multi-graph matching method, which not only avoids the cycle-consistency problem, but also reduces the complexity. In addition, we further improved the proposed method by introducing a lossless factorization of the affinity matrix in the multi-graph matching methods. Synthetic and real data experiments demonstrate the superiority of our method.

High Performance Time Series Anomaly Detection Using Brain Inspired Cortical Coding Method

Accurate and automated anomaly detection in time series data sets has an increasingly important role in a wide range of applications. Inspired by coding in the cortical networks of the brain, here we introduce a novel approach for high performance real-time anomaly detection. Cortical coding method is adaptive and dynamic, consisting of self-organized networks. In the cortical coding network introduced herein, the morphological structuring is driven by a brain inspired feature extraction strategy that aims the minimization of the signal energy dissipation while increasing the information entropy of the system. We combine the cortical coding network with transform coding and multi resolution analysis for anomaly detection. As we demonstrate here, the new coding methodology provides high computational efficiency in addition to scalability with respect to target accuracy compared to the traditional clustering algorithms. A wide variety of data sets are used to demonstrate time series anomaly detection performance. In a preliminary work presented here, we detected 77.6% of the present anomalies correctly, using the same hyperparameters for every stage of the method. The results are compared with several clustering algorithms such as K-means and its variants mini-batch K-means, sequential K-means and finally with hierarchical agglomerative clustering. Additionally, the performance of all the clustering methods are compared by memorizing all input data set without performing any clustering. The cortical coding method has shown the best performance compared to the other methods. From the results achieved so far, it appears that there is still a significant room for improvement of the success rate by, specifically, performing hyperparameter and filter optimization according to the characteristics of data sets and using a more advanced fusion model at the output layer. Low time and space complexity, high generalization performance, suitability to real-time anomaly detection, and in-memory processing compatibility are distinct advantages of the cortical coding method in a variety of anomaly detection problems, such as predictive maintenance, cybersecurity, telemedicine, risk management, and transportation safety.

A High Phase Detection Density and Low Space Complexity Mueller-Muller Phase Detector for DB PAM-4 Wireline Receiver

A Mueller-Muller Phase Detector (MM PD) technology based on duo-binary four-level pulse amplitude modulation (DB PAM-4) with low complexity and high phase-detection density is presented. The proposed low complexity includes low phase-detection complexity and low space complexity of data processing. The waveform sifting technology simplifies 175 specific waveform changes into five fuzzy waveform change trends, reducing the complexity of subsequent phase detection. By making the data sample before the waveform sifting, the data bit width is reduced from 8 bit to 3 bit, which realizes data dimensionality reduction, greatly reduces the scale of subsequent auxiliary data, reduces the number of basic devices by 13.7%, and reduces the spatial complexity of data processing. The coherent coding of DB PAM-4 combined with waveform sifting increases the phase-detection density from 50% to 65% and improves both phase-detection density and phase-detection gain by 30%, and improves the jitter tolerance. Through the simulation of the clock and data recovery (CDR) model built by Cadence, the fast locking capability of CDR is verified.

Clover: tree structure-based efficient DNA clustering for DNA-based data storage.

Deoxyribonucleic acid (DNA)-based data storage is a promising new storage technology which has the advantage of high storage capacity and long storage time compared with traditional storage media. However, the synthesis and sequencing process of DNA can randomly generate many types of errors, which makes it more difficult to cluster DNA sequences to recover DNA information. Currently, the available DNA clustering algorithms are targeted at DNA sequences in the biological domain, which not only cannot adapt to the characteristics of sequences in DNA storage, but also tend to be unacceptably time-consuming for billions of DNA sequences in DNA storage. In this paper, we propose an efficient DNA clustering method termed Clover for DNA storage with linear computational complexity and low memory. Clover avoids the computation of the Levenshtein distance by using a tree structure for interval-specific retrieval. We argue through theoretical proofs that Clover has standard linear computational complexity, low space complexity, etc. Experiments show that our method can cluster 10 million DNA sequences into 50 000 classes in 10 s and meet an accuracy rate of over 99%. Furthermore, we have successfully completed an unprecedented clustering of 10 billion DNA data on a single home computer and the time consumption still satisfies the linear relationship. Clover is freely available at https://github.com/Guanjinqu/Clover.

Hybrid Maximum Clique Algorithm Using Parallel Integer Programming for Uniform Test Assembly

Educational assessments often require uniform test forms, for which each test form has equivalent measurement accuracy but with a different set of items. For uniform test assembly, an important issue is the increase of the number of assembled uniform tests. Although many automatic uniform test assembly methods exist, the maximum clique algorithm (MCA)-based method is known to assemble the greatest number of uniform tests with the highest measurement accuracy based on the item response theory. In that method, the graph is constructed by sequentially adding a randomly formed test as a vertex without considering the graph structure. However, an important difficulty is its high space complexity, which interrupts search cliques with more than a hundred thousand vertices. To overcome this difficulty, this article proposes a new uniform test assembly algorithm: hybrid maximum clique algorithm using parallel integer programming. The first step searches a maximum clique that is as large as possible up to computer memory limitations using a state-of-the-art MCA with low time complexity but with high space complexity. The second step repeatedly searches a vertex connected with all vertices of the current maximum clique from the remaining vertices using integer programming with low space complexity but with high time complexity. The proposed method constructs a larger number of tests than the traditional methods do. Finally, we use simulation and actual data experiments to demonstrate the effectiveness of the proposed method. Results show that our method assembles a 1.5–2.7 times greater number of uniform tests than traditional methods can.

HashWalk: An efficient node classification method based on clique-compressed graph embedding

• Based on homophily and co-citation hypothesis, cliques reflect structural features of a graph. • Considering a clique as a sample unit improves the quality of random walk . • Compressing graph reduces computational cost in graph embedding. • Compressing graph by consistent cliques preserves structural information. In recent years, random walk based embedding has become a popular method for node classification. However, current methods still require a huge computational cost to obtain the representation of a large number of nodes. In addition, walking methods cannot adapt well to diverse network structures. Hence, this paper proposes HashWalk to generate a clique-compressed graph that can be used in random walk based embedding for node classification. Specifically, HashWalk compresses cliques into single nodes, and these single nodes are able to inherit neighbors of cliques. As a result, HashWalk can significantly reduce the number of training nodes and computational cost. Besides, HashWalk uses the random walk mapping method to obtain walking sequences of the clique-compressed graph, which makes random walk adapt to network structure. The experimental results prove that HashWalk provides faster time efficiency and lower space complexity while ensuring the accuracy. In summary, this paper provides a fast and effective method using clique-compressed graph embedding for node classification.

Deep learning-based fast detection of apparent concrete crack in slab tracks with dilated convolution

Slab tracks exposed to complicated environmental factors over a long period can cause cracks in the concrete, and if these cracks gradually expand, the concrete’s durability and service life will be greatly impacted. How to quickly and effectively detect concrete cracks has become an urgent challenge during the maintenance and repair of high-speed railway slab tracks. In this study, a large number of images of concrete cracks were collected in a database, and STCNet Ⅰ, a fast detection network architecture using dilated convolution based on deep learning, was proposed to detect apparent concrete cracks in slab tracks. After that, the watershed algorithm was used to segment the detected cracks. The results show that: I) compared with traditional network models, the STCNet Ⅰ provides a faster calculation at lower space complexity. The number of parameters used in this network is reduced by 96.03% and 93.28%, respectively compared with that in the VGG 16 and ResNet 50, and the time complexity is lower, with the calculation time reduced by 49.94% and 73.28%, respectively; II) the average recognition accuracy on the training set and the validation set reached as high as 99.71% and 99.33%, respectively, proving the robustness of the model; III) the accuracy and F1 score in the test samples of concrete crack reached 99.54% and 99.54%, indicating the strong generalization ability of the model; and IV) the concrete crack area was accurately detected, and the crack contour was fully closed and continuous. The research results from this paper provide an improved detection method of slab tracks and promote the fine detection and maintenance of the apparent concrete of slab tracks.

Deep Expectation-Maximization for Joint MIMO Channel Estimation and Signal Detection

To overcome the influence of channel estimation error on signal detection, this paper presents a model-driven deep learning method for joint channel estimation and signal detection in multiple-input multiple-output (MIMO) wireless communication systems. To improve the robustness of the method, we use the Student's <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">t</i> -distribution to model the environment noise, and model the joint problem probabilistically by taking the channel state information (CSI) as a latent variable, then derive a generalized expectation maximization (GEM) algorithm to fit the model. In GEM, the expectation step is used for robust CSI estimation while the maximization step detects the transmitted signal with the estimated CSI. To reduce the computational complexity of GEM, we unfold it into a deep neural network. The network contains only a few trainable parameters, which is easy to train and has low space complexity. Based on our experimental results, we find that the proposed GEM algorithm outperforms a variety of signal detection algorithms which take the CSI estimation and signal detection as independent modules. Further, we find empirically that the proposed GEM network outperforms state-of-the-art deep learning based joint channel estimation and signal detection methods and has a good generalization ability against the number of antennas, the modulation mode, the level of signal-to-noise ratio and channel correlation.