Concatenation Structure Research Articles

DFW-YOLO: YOLOv5-based algorithm using phased array ultrasonic testing for weld defect recognition

ABSTRACT Although phased array ultrasonic testing (PAUT) technology has been widely employed in pipeline weld defect recognition, current defect recognition still relies on manual operation, which limits its efficiency. By combining deep learning with PAUT, this paper presents a new neural architecture based on YOLOv5 called DFW-YOLO that can call defect indications automatically. To address redundant defect feature maps, the FasterNet backbone is introduced to enhance the model’s feature extraction ability. Additionally, partial dilated convolution (PDConv) and partial convolution (PConv) with an atrous convolution branch are designed to extend the receptive field of the backbone network, mitigating the chance of feature loss. Furthermore, a weighted generalised feature pyramid network (WGFPN) structure is proposed as the neck network. This structure combines the reparameterized generalised feature pyramid network (RepGFPN) with the weighted aggregated concatenation (WAC) structure and eliminates inefficient branches to enhance the model’s feature fusion ability. The experimental results demonstrated that DFW-YOLO outperforms the original YOLOv5 in terms of detection performance. The precision, recall, mean average precision (mAP)@0.5, and mAP@0.5:0.95 are indicators for DFW-YOLO and reach 97.4%, 97.5%, 98.3%, and 68.2%, respectively, marking 4.8%, 5%, 2.7%, and 22.5% improvement compared to the original YOLOv5.

A Spiking Neural Network Based on Retinal Ganglion Cells for Automatic Burn Image Segmentation.

Burn is a common traumatic disease. After severe burn injury, the human body will increase catabolism, and burn wounds lead to a large amount of body fluid loss, with a high mortality rate. Therefore, in the early treatment for burn patients, it is essential to calculate the patient's water requirement based on the percentage of the burn wound area in the total body surface area (TBSA%). However, burn wounds are so complex that there is observer variability by the clinicians, making it challenging to locate the burn wounds accurately. Therefore, an objective, accurate location method of burn wounds is very necessary and meaningful. Convolutional neural networks (CNNs) provide feasible means for this requirement. However, although the CNNs continue to improve the accuracy in the semantic segmentation task, they are often limited by the computing resources of edge hardware. For this purpose, a lightweight burn wounds segmentation model is required. In our work, we constructed a burn image dataset and proposed a U-type spiking neural networks (SNNs) based on retinal ganglion cells (RGC) for segmenting burn and non-burn areas. Moreover, a module with cross-layer skip concatenation structure was introduced. Experimental results showed that the pixel accuracy of the proposed reached 92.89%, and our network parameter only needed 16.6 Mbytes. The results showed our model achieved remarkable accuracy while achieving edge hardware affinity.

Adaptive Belief Propagation Decoding of CRC Concatenated NR LDPC and Polar Codes

In this paper, we propose a modified adaptive belief propagation (ABP) algorithm, which is referred to as the random ABP (R-ABP) algorithm, for decoding short channel codes adopted in the fifth generation (5G) New Radio (NR) wireless systems. Based on the cyclic redundancy check (CRC) concatenation structure of 5G channel codes, we can take partial or even all CRC bits into iterative R-ABP decoding to improve the error correction performance, while the error detection capability can still be guaranteed by the remaining CRC bits and a proposed threshold-based acceptance criterion. We call this improved R-ABP decoding the threshold-and-CRC-aided R-ABP (TCA-R-ABP) decoding. The simulation results show that our proposed TCA-R-ABP algorithm outperforms the state-of-the-art decoding algorithms for many 5G low-density parity-check (LDPC) and polar codes. Moreover, the proposed TCA-R-ABP algorithm leads to a unified decoder for LDPC and polar codes, which has the potential to be a lower-complexity decoder over the combination of the belief propagation (BP) and CRC-aided successive cancellation list (CA-SCL) decoders.

The Structure of Optimal Protocols for a Mathematical Model of Chemotherapy with Antiangiogenic Effects

We analyze a mathematical model for cancer chemotherapy which includes antiangiogenic effects of the cytotoxic agent. Assuming that the total amount of agents to be given has been determined a priori based on a medical assessment of its side effects, we consider the problem how to best administer this amount. The model assumes a homogenous tumor and if the aim is to minimize the tumor volume, then optimal controls administer the total dose in a single maximum dose session. As, however, angiogenic effects of the agent are taken into account, this no longer is optimal. Lower dose strategies determined by an optimal singular arc with significantly reduced dose rates give a better response over time. In this paper, for the medically realistic domain of the mathematical model, the concatenation structure of optimal controlled trajectories as segments of bang and singular arcs is determined. This leads to simple numerical minimization procedures for the computation of globally optimal controls.

Strong Noise-Tolerance Deep Learning Network for Automatic Microseismic Events Classification

Identifying useful microseismic events is one of the key steps in monitoring tunnel rockbursts. Here, we propose a strong noise-tolerance deep learning (SNTDL) network for the automatic classification of noisy microseismic events. The training set, validation set, and test set of the SNTDL network consist of 27,989 unfiltered microseismic recordings. First, to comprehensively characterize the microseismic events, we extract 10 weakly correlated features of the microseismic recordings as the input of the SNTDL network. Then, the skip connection and concatenation structure are added to this network, which can further enhances its generalization ability. Additionally, the SNTDL, AlexNet, Inception, Visual Geometry Group, and ResNet are compared using the synthetic microseismic recordings with different signal-noise ratios. The results demonstrate that the SNTDL network has a higher accuracy and stronger noise-tolerance capability than the other approaches. Application to a dataset collected from a different construction environment confirms that the SNTDL network can still achieve an accurate classification result, which further verifies that the proposed network has a reliable generalization performance.

Single image dehazing algorithm based on pyramid mutil-scale transposed convolutional network

In this paper, a novel single image dehazing method based on pyramid multi-scale transposed convolutional network (MST-Net) is proposed. Conventional haze removal algorithms based on the atmospheric scattering model may lead to some problems such as incomplete dehazing and colour distortion due to the inaccurate parameter approximations or indirect image optimization and reconstruction. Therefore, we design a real end-to-end image dehazing network to directly learn the mapping relationship between hazy images and the corresponding clear images. In this network, the cascaded feature extraction blocks extract the diversified feature information of the input images by multi-channel concatenation structure, which can effectively fuse the local features of the first convolution layer into the semantic features of subsequent layers in the block. In order to reconstruct high-quality dehazed images relieving the colour distortion, we design a multi-scale transposed convolution block to gradually expand the resolution of the obtained feature maps, and introduce skip connections from the feature extraction module to supplement the detailed information of the feature map pyramid. Extensive experimental results demonstrate that the proposed method in this paper can remove the haze completely and achieve superior performance in subjective and objective evaluation over the other state-of-the-art methods.

Text Detection and Recognition for Images of Medical Laboratory Reports With a Deep Learning Approach

The adoption of electronic health records (EHRs) is an important step in the development of modern medicine. However, complete health records are not often available during treatment because of the functional problem of the EHR system or information barriers. This paper presents a deep-learning-based approach for textual information extraction from images of medical laboratory reports, which may help physicians solve the data-sharing problem. The approach consists of two modules: text detection and recognition. In text detection, a patch-based training strategy is applied, which can achieve the recall of 99.5% in the experiments. For text recognition, a concatenation structure is designed to combine the features from both shallow and deep layers in neural networks. The experimental results demonstrate that the text recognizer in our approach can improve the accuracy of multi-lingual text recognition. The approach will be beneficial for integrating historical health records and engaging patients in their own health care.

Skew multi-twisted codes over finite fields and their Galois duals

In this paper, we introduce a new class of linear codes over finite fields, viz. skew multi-twisted codes (or skew generalized quasi-twisted codes), which is a generalization of some well-known classes of linear codes such as cyclic codes and quasi-cyclic codes. We also study algebraic structures of skew multi-twisted codes and their Galois duals (i.e., orthogonal complements with respect to the Galois inner product). We also view skew multi-twisted codes as direct sums of certain concatenated codes, which gives rise to a method to construct these codes. We also determine a lower bound on their minimum Hamming distances using their multilevel concatenated structure. Apart from this, we determine the parity-check polynomial of each skew multi-twisted code, and obtain BCH type bounds on their minimum Hamming distances. We also determine generating sets of Galois duals of some skew multi-twisted codes from generating sets of these codes. We also derive necessary and sufficient conditions under which a skew multi-twisted code is (i) Galois self-dual, (ii) Galois self-orthogonal and (iii) Galois LCD (linear with complementary dual). As an application of this result, we also provide enumeration formulae for all Hermitian self-dual, Hermitian self-orthogonal, Hermitian LCD and Euclidean LCD multi-twisted codes over finite fields. We also obtain many linear codes with best known and optimal parameters from 1-generator skew multi-twisted codes over finite fields F8 and F9.

CRC Location Design for Polar Codes

Polar codes concatenated with cyclic redundancy check (CRC) can achieve a competitive performance compared with the state-of-the-art codes. In this letter, we propose a criterion for designing concatenation structure based on an error pattern matrix. According to this criterion, we rearrange the location of CRC bits for polar codes to further improve the decoding performance. Meanwhile, the optimal CRC searching complexity of the proposed concatenation structure is significantly less than that of the standard CRC-polar concatenation scheme. Simulation results show that the proposed concatenation scheme outperforms the standard scheme whether they use the identical standard CRC or their respective optimal CRCs.

Multi-twisted codes over finite fields and their dual codes

Let Fq denote the finite field of order q, let m1,m2,⋯,mℓ be positive integers satisfying gcd⁡(mi,q)=1 for 1≤i≤ℓ, and let n=m1+m2+⋯+mℓ. Let Λ=(λ1,λ2,⋯,λℓ) be fixed, where λ1,λ2,⋯,λℓ are non-zero elements of Fq. In this paper, we study the algebraic structure of Λ-multi-twisted codes of length n over Fq and their dual codes with respect to the standard inner product on Fqn. We provide necessary and sufficient conditions for the existence of a self-dual Λ-multi-twisted code of length n over Fq, and obtain enumeration formulae for all self-dual and self-orthogonal Λ-multi-twisted codes of length n over Fq. We also derive some sufficient conditions under which a Λ-multi-twisted code is linear with complementary dual (LCD). We determine the parity-check polynomial of all Λ-multi-twisted codes of length n over Fq and obtain a BCH type bound on their minimum Hamming distances. We also determine generating sets of dual codes of some Λ-multi-twisted codes of length n over Fq from the generating sets of the codes. Besides this, we provide a trace description for all Λ-multi-twisted codes of length n over Fq by viewing these codes as direct sums of certain concatenated codes, which leads to a method to construct these codes. We also obtain a lower bound on their minimum Hamming distances using their multilevel concatenated structure.

Practical Design and Decoding of Parallel Concatenated Structure for Systematic Polar Codes

In this paper, a parallel concatenated structure is proposed to improve the performance of polar codes in finite length in which multiblocks systematic polar codes (SPC) and recursive systematic convolutional codes (RSC) are combined. We first design a weighted iterative decoding algorithm for the proposed structure such that better bit error rate performance can be obtained. Then, a deliberated interleaver is optimized for the new scheme. Thereafter, we introduce a strategy to estimate the minimum distance of the concatenation codeword, where an asymptotic performance and convergence behavior analysis are elaborated. Finally, the error performance of the proposed scheme is evaluated via simulations. The results show that our scheme almost outperforms all the existed concatenation ones using polar codes, making the codes more practical.

Design and Construction of a Historical Financial Database of the Italian Stock Market 1973--2011

This article presents the technical aspects of designing and building a historical database of the Italian Stock Market. The database contains daily market data from 1973 to 2011 and is constructed by merging two main digital sources and several other hand-collected data sources. We analyzed and developed semiautomatic tools to deal with problems related to time-series matchings, quality of data, and numerical errors. We also developed a concatenation structure to allow the handling of company name changes, mergers, and spin-offs without artificially altering numerical series. At the same time, we maintained the transparency of the historical information on each individual company listed. Thanks to the overlapping of digital and hand-collected data, the completed database has a very high level of detail and accuracy. The dataset is particularly suited for any empirical research in financial economics and for more practically oriented numerical applications and forecasting simulations.

Generalization of extensive structures and its representation

This paper generalizes extensive structures so that a weighted additive model can be obtained. A left nonnegative concatenation structure with left identity is defined as a nonnegative concatenation structure (Luce et al., 1990) with left identity for which the solvability and Archimedean properties are satisfied only related to left-concatenation. This structure has two partial binary operations–multiplication and right division–and a new partial binary operation is defined on it. Two conditions of equivalence form are then provided to make the left nonnegative concatenation structure with left identity into an extensive structure with identity with respect to the newly defined operation. Finally, the weighted additive model is derived from an additive representation on the extensive structure, so that distinct m-period and n-period (m≠n) temporal sequences can be compared.

A survey of three-dimensional turbo codes and recent performance enhancements

This paper presents a survey of two techniques intended for improving the performance of conventional turbo codes (TCs). The first part of this work is dedicated to explore a hybrid concatenation structure combining both parallel and serial concatenation based on a three-dimensional (3D) code. The 3D structure, recently introduced by Berrou et al., is able to ensure large asymptotic gains at very low error rates at the expense of an increase in complexity and a loss in the convergence threshold. In order to reduce the loss in the convergence threshold, the authors consider first a time-varying construction of the post-encoded parity. Then, they investigate the association of the 3D TC with high-order modulations according to the bit-interleaved coded modulation approach. The second part of this study deals with irregular TCs. In contrast to 3D TCs, although irregular TCs can achieve performance closer to capacity, their asymptotic performance is very poor. Therefore, the authors propose irregular turbo coding schemes with suitable interleavers in order to improve their distance properties. Finally, a modified encoding procedure, inspired from the 3D TC, makes it possible to obtain irregular TCs which perform better than the corresponding regular codes in both the waterfall and the error floor regions.

Generalized factorizable automorphisms in n-component conjoint structures

The group of generalized factorizable automorphisms (Math. Soc. Sci. 36 (1998) 91) of a n-component conjoint structure, which satisfies certain solvability assumptions and the Thomsen condition, is proved to be the semidirect product of its normal subgroup of factorizable automorphisms and a subgroup isomorphic to the symmetric group of degree n. This sets the frame for discussing how solvability assumptions are mirrored by properties of automorphisms, for deriving results on the existence of factorizable automorphisms, and for relating the generalized factorizable automorphisms to the isomorphisms of the concatenation structures induced on the components of a conjoint structure. It is concluded that the generalized factorizable automorphisms do not provide any contribution to the independence of factors of a n-component conjoint structure other than that already captured by the group of factorizable automorphisms. The results do not draw upon Archimedean axioms.

Singular points in generalized concatenation structures that otherwise are homogeneous

A generalized concatenation structure X is any ordered relational structure in which at least one relation is a function of two or more variables that is monotonically increasing in each of them. A point of X is said to be singular if it is fixed under all automorphisms of the structure. Examples are 0 in the additive real numbers, the velocity of light in special relativity, and the status quo in some theories of utility. A translation is an automorphism with no fixed points other than the singular ones. It is assumed that X is order dense and finitely unique, from which it follows (Theorem 1) that at most three singular points may exist; a maximum, a minimum, and/or an interior one. Assuming also that X is homogeneous in translations between adjacent singularities, the singular points are further characterized (Theorem 2), with the interior one being shown to be a generalized zero (Definition 4). It is shown how to replace a generalized zero by a zero (Theorem 3) and how to partition X into homogeneous structures on either side of the interior generalized zero (Theorem 4). The major problem in representing such structures lies in understanding how the two halves relate. For structures on a continuum that are translation homogeneous between singularities, finitely unique, and solvable relative to the zero, a representation exists in terms of unit structures (Theorem 5).

Rank- and sign-dependent linear utility models for finite first-order gambles

Finite first-order gambles are axiomatized. The representation combines features of prospect and rank-dependent theories. What is novel are distinctions between gains and losses and the inclusion of a binary operation of joint receipt. In addition to many of the usual structural and rationality axioms, joint receipt forms an ordered concatenation structure with special features for gains and losses. Pfanzagl's (1959) consistency principle is assumed for gains and losses separately. The nonrational assumption is that a gamble of gains and losses is indifferent to the joint receipt of its gains pitted against the status quo and of its losses against the status quo.

Dedekind completion of Positive Concatenation Structures: Necessary and sufficient conditions

Necessary and sufficient conditions are presented to Dedekind complete a Closed Positive Concatenation Structure. In addition, a new graphical method is presented for verifying the axioms of a Positive Concatenation Structure, and a counter-example is exhibited to the claim that all Positive Concatenation Structures may be Dedekind completed.

Measurement structures with archimedean ordered translation groups

The paper focuses on three problems of generalizing properties of concatenation structures (ordered structures with a monotonic operation) to ordered structures lacking any operation. (1) What is the natural generalization of the idea of Archimedeaness, of commensurability between large and small? (2) What is the natural generalization of the concept of a unit concatenation structure in which the translations (automorphisms with no fixed point) can be represented by multiplication by a constant? (3) What is the natural generalization of a ratio scale concatenation structure being distributive in a conjoint one, which has been shown to force a multiplicative representation of the latter and the product-of-powers representation of units found in physics? It is established (Theorems 5.1 and 5.2) that for homogeneous structures, the latter two questions are equivalent to it having the property that the set of all translations forms a homogeneous Archimedean ordered group. A sufficient condition for Archimedeaness of the translations is that they form a group, which is equivalent to their being 1-point unique, and the structure be Dedekind complete and order dense (Theorems 2.1 and 2.2). It is suggested that Archimedean order of the translations is, indeed, also the answer to the first question. As a lead into that conclusion, a number of results are reported in Section 3 on Archimedeaness in concatenation structures, including for positive structures sufficient conditions for several different notions of Archimedeaness to be equivalent. The results about idempotent structures are fragmentary.

Uniqueness and homogeneity of ordered relational structures

There are four major results in the paper. (1) In a general ordered relational structure that is order dense, Dedekind complete, and whose dilations (automorphisms with fixed points) are Archimedean, various consequences of finite uniqueness are developed (Theorem 2.6). (2) Replacing the Archimedean assumption by the assumption that there is a homogeneous subgroup of automorphisms that is Archimedean ordered is sufficient to show that the structure can be represented numberically as a generalized unit structure in the sense that the defining real relations satisfy the usual numerical property of homogeneity (Theorem 3.4). The last two results pertain just to idempotent concatenation structures. (3) In a closed, idempotent, solvable, and Dedekind complete concatenation structure, homogeneity is equivalent to the structure satisfying an inductive property analogous to the condition for homogeneity in a positive concatenation structure (Theorem 4.3). Finally, (4) an axiomatization is given for an idempotent structure to be of scale type (2, 2), which has previously been shown to be equivalent to a dual bilinear representation. Basically two operations are defined in terms of the given one, and the conditions are that each must be right autodistributive and together they satisfy a generalized bisymmetry property. The paper ends listing several unsolved problems.