Bidirectional Structures Research Articles

Optimizing hybrid neural networks for precise COVID-19 mRNA vaccine degradation prediction

Conventional hybrid models often miss an essential factor that can lead to less effective performance: intrinsic sequence dependence when combining various neural network (NN) architectures. This study addresses this issue by highlighting the importance of sequence hybridization in NN architecture integration, aiming to improve model effectiveness. It combines NN layers—dense, long short-term memory (LSTM), and gated recurrent unit (GRU)—using the Keras Sequential API for defining the architecture. To provide better context, bidirectional LSTM (BiLSTM) and bidirectional GRU (BiGRU) replace their unidirectional counterparts, enhancing the models through bidirectional structures. Out of 25 NN models tested, 18 four-layer hybrid NN models consist of one-quarter dense layer and the rest BiLSTM and BiGRU layers. These hybrid NN models undergo supervised learning regression analysis, with mean column-wise root mean square error (MCRMSE) as the performance metric. The results show that each hybrid NN model produces unique outcomes based on its specific hybrid sequence. The Hybrid_LGSS model performs better than existing three-layer BiLSTM networks in predictive accuracy and shows lower overfitting (MCRMSEs of 0.0749 and 0.0767 for training and validation, respectively). This indicates that the optimal hybridization sequence is crucial for achieving a balance between performance and simplicity. In summary, this research could help vaccinologists develop better mRNA vaccines and provide data analysts with new insights for improvement.

COMPARISON OF DEEP LEARNING TECHNIQUES FOR TEXTUAL SENTIMENT ANALYSIS

Examining the opinions of the people might provide us with valuable knowledge. Sentiment analysis is a method for analyzing textual data that helps find subjective information, such as opinions and feelings that individuals or groups have expressed. It improves our understanding of human language using deep learning and natural language techniques. Several deep learning models, including RNNs, LSTMs, GRUs, and their bidirectional variants, are compared in this work. Three publicly available datasets - the imdb_reviews, Twitter Sentiment Dataset, and Emotions dataset were used in the investigation. Accuracy performance is evaluated for six deep learning models. According to experimental studies, bidirectional structures outperform their unidirectional counterparts in most cases. Across several datasets, the bidirectional models continuously produced the best accuracy.

On the Design of Continuously Variable Transmissions with Bidirectional Bridge Structures for Hybrid Vehicles

To increase the energy efficiency of road vehicles, an ideal transmission system should have a wide ratio coverage, a high torque capacity, and a high mechanical efficiency. Continuously variable units (CVUs) have been successfully implemented due to the smooth ratio variation, sufficient torque capacity, and ratio coverage. Hence, it will be beneficial to develop a hybrid powertrain comprising a CVU. In this paper, a design method called the “basic path diagram” (BPD) is proposed. It provides a simplified schematic of the system and represents the generic connections among the mechanical components. The system configurations synthesized by the BPD can be sorted according to three characteristics: Direction of power flows through the CVU, coupling pattern of the power inputs, and number of transmission paths parallel to the CVU. The first characteristic determines the number of times the CVU ratio coverage can be exploited, the second characteristic determines whether the torque of the power inputs can be independently controlled, and the third characteristic can help reduce the torque loading of the CVU. With the aid of a BPD, one of the possible system configurations is provided as an example. The result shows that the system can exploit twice the ratio coverage of the CVU and reduce the torque and power transmitted by the CVU in combination with planetary gearsets.

A semi-analytical mesh-free method for 3D free vibration analysis of bi-directional FGP circular structures subjected to temperature variation

In this present paper, a semi-analytical mesh-free method is employed for the three-dimensional free vibration analysis of a bi-directional functionally graded piezoelectric circular structure. The dependent variables have been expanded by Fourier series with respect to the circumferential direction and have been discretized through radial and axial directions based on the mesh-free shape function. The current approach has a distinct advantage. The nonlinear Green-Lagrange strain is employed as the relationship between strain and displacement fields to observe thermal impacts in stiffness matrices. Nevertheless, high order terms have been neglected at the final steps of equations driving. The material properties are assumed to vary continuously in both radial and axial directions simultaneously in accordance with a power law distribution. The convergence and validation studies are conducted by comparing our proposed solution with available published results to investigate the accuracy and efficiency of our approach. After the validation study, a parametric study is undertaken to investigate the temperature effects, different types of polarization, mechanical and electric boundary conditions and geometry parameters of structures on the natural frequencies of functionally graded piezoelectric circular structures.

Investigating the use of uni-directional and bi-directional long short-term memory models for automatic sleep stage scoring

Abstract In this paper, a study is conducted to investigate the use of a Long Short-Term Memory (LSTM) learning system in automatic sleep stage scoring. The developed algorithm will automatically learn to classify sleep stages from any acquired sleep signals data-set. This allows to resolve the difficulties that are facing experts in manual sleep stage scoring. A total of 39 Polysomnogram (PSG) recordings acquired from the online PhysioNet Sleep-EDF database are used in this study. The PSG recordings are chosen to be from the EEG Fpz-Cz signals only. The database comes with annotation files that include expert manual stage scoring based on the Rechtschaffen & Kales (R & K) scoring manual. The obtained signals go initially through a pre-processing procedure where sleep stages signals are extracted, normalized, and filtered. The resulting sleep signals are trained using a k-fold cross-validation scheme of 10-folds. Prior to the training and classification process, the LSTM network architecture is built using Uni- and Bi-directional structures to utilize both the forward and backward chains of data sequences. At the end, the developed algorithm performance is evaluated and a complete performance summary table is provided relative to other State-of-the-Art deep learning studies. The performance of this study is evaluated initially without the merging of S3 and S4 sleep stages following the R & K manual, which is considered challenging due to the minor differences between the signals. Then, the performance is evaluated following the recent American Academy of Sleep Medicine (AASM) scoring manual with the merging of the two stages as N3. The developed algorithm achieved higher results using the Bi-directional LSTM. In addition, it achieved the highest accuracy among all other studies in the field with 97.28 % . Furthermore, Cohen's kappa and F1-score were more than 72 % on average between all sleep stages. According to the confusion matrix, the algorithm successfully classified sleep signals with an overall True Positives percentage of 91.92 % . The performance of the algorithm improved following the AASM manual, where the Cohen's kappa value increased from 72.55 % to 77.73 % . The developed algorithm showed potential in automatic sleep stage classification. Future works include further enhancements on the LSTM algorithm to achieve higher levels of performance.

Non‐isolated interleaved bidirectional DC–DC converter with high step voltage ratio and minimum number of switches

In this study, a new non-isolated interleaved bidirectional DC–DC converter with a high step voltage ratio based on coupled inductors (CIs) is introduced. The CIs significantly increase the voltage conversion ratio and provide soft-switching conditions for the switches at turn-on instant. Also, employing an interleaved structure combined with the CI in the low-voltage side noticeably reduces the current ripple and this feature has a vital role in the high-power applications. Also, this topology provides auto-current sharing for the two phases of the interleaved structure. The CIs in the high-voltage side, supply a steep gain in both buck and boost modes with an appropriate duty cycle. The proposed method provides zero current switching (ZCS) turn-off condition for all diodes and ZCS turn-on condition for all switches without requiring any auxiliary circuit. So the undesirable reverse recovery effects of diodes are considerably reduced. The proposed converter has the minimum number of switches for two-phase bidirectional structures. Also, a simple passive clamp circuit is proposed for the converter to limit the voltage spikes of CIs. Finally, a 600 W, 30/300 V laboratory prototype is implemented to verify the performance.

Tidal bar cyclicity record in the Lower Permian: The Rio Bonito Formation, Paraná Basin, southern Brazil

Tidal bars register gravitational interactions between the Earth, Moon, and Sun and allow inferences about astronomical positioning through the cyclic variation of sandstone layer thickness throughout depositional events. Measurements of sandstone layers thickness in a tidal bar of the Rio Bonito Formation (Paraná Basin, southern Brazil) and further spectral analysis allowed the definition of the tide patterns imprinted on rock layers. The study was based on a remarkable tidal bar outcrop located in the Candiota paleovalley in the southern basin, Candiota town, Brazil, with sedimentary characteristics that suggest to a tidally dominated estuary depositional environment. The tidal bar shows a medium- to large-scale cross-stratification, reactivation surfaces, bidirectional structures, and tidal bundles. The spectral analysis resulted in a significant frequency peak (0.0329) which corresponds to a period of 30.45 layers per cycle. Six of these cycles are recognizable in the tidal bar by its cyclic thickening-thinning of layers, that range from 1.5 to 11 cm. This cyclic pattern is interpreted as being produced by the neap-spring tide cycle. Our study also allows us to infer that the south Paraná Basin Permian tide had a semi-diurnal pattern, with a neap-spring tide period of 14.9 days. Comparison with previous studies points to small changes in the duration of the neap-spring tide cycle in the last 900 Ma (15.2 days in the Cambrian, 14.9 days in the early Permian and 14.7 days today), explained by the distance and acceleration of the movement of the Moon's revolution and decrease in the speed of rotation of the Earth.

A novel wavelet sequence based on deep bidirectional LSTM network model for ECG signal classification

Long-short term memory networks (LSTMs), which have recently emerged in sequential data analysis, are the most widely used type of recurrent neural networks (RNNs) architecture. Progress on the topic of deep learning includes successful adaptations of deep versions of these architectures. In this study, a new model for deep bidirectional LSTM network-based wavelet sequences called DBLSTM-WS was proposed for classifying electrocardiogram (ECG) signals. For this purpose, a new wavelet-based layer is implemented to generate ECG signal sequences. The ECG signals were decomposed into frequency sub-bands at different scales in this layer. These sub-bands are used as sequences for the input of LSTM networks. New network models that include unidirectional (ULSTM) and bidirectional (BLSTM) structures are designed for performance comparisons. Experimental studies have been performed for five different types of heartbeats obtained from the MIT-BIH arrhythmia database. These five types are Normal Sinus Rhythm (NSR), Ventricular Premature Contraction (VPC), Paced Beat (PB), Left Bundle Branch Block (LBBB), and Right Bundle Branch Block (RBBB). The results show that the DBLSTM-WS model gives a high recognition performance of 99.39%. It has been observed that the wavelet-based layer proposed in the study significantly improves the recognition performance of conventional networks. This proposed network structure is an important approach that can be applied to similar signal processing problems.

Inelastic Parametric Analysis of Seismic Responses of Multistorey Bidirectional Eccentric Structure

This paper conducts a parametric study on the seismic response of multistorey bidirectional eccentric structures from elastic stage to inelastic stage. Based on a simplified multistorey bidirectional eccentric model composed of bidirectional lateral load‐resisting members, a general law is proposed for three‐stage natural frequency variation behaviour from elastic stage to inelastic stage of eccentric frame structures with different layers. Different simplification treatments are conducted on each stage and the three stable parameter analysis stages are defined. The corresponding dynamic stiffness matrices and motion equations in different loading stages are derived. On this basis, a parametric analysis of seismic response of a three‐storey bidirectional regular eccentric structure from elastic stage to inelastic stage is conducted. Effects of the uncoupled torsion to lateral frequency ratios (Ω) and bidirectional eccentricities on the seismic responses are investigated. The results reveal that as Ω increases, translational displacement in the load direction first decreases and then increases; meanwhile, the displacement perpendicular to load direction and torsion displacement first rise and then decrease sharply. When Ω = 1.1, the coupling effect between the translation in the load direction and the torsion is at its strongest condition. Increasing the eccentricities leads to a decrease in the displacement in the load direction as well as an increase in the displacement perpendicular to load direction and torsion displacement. Variation regularity of inelastic seismic response is remarkably different from that in elastic stage. The lateral‐torsional coupling effect of the bidirectional eccentric structure is closely related to both the period ratio and the bidirectional eccentricities.

Nonrandom network connectivity comes in pairs

Overrepresentation of bidirectional connections in local cortical networks has been repeatedly reported and is a focus of the ongoing discussion of nonrandom connectivity. Here we show in a brief mathematical analysis that in a network in which connection probabilities are symmetric in pairs, Pij = Pji, the occurrences of bidirectional connections and nonrandom structures are inherently linked; an overabundance of reciprocally connected pairs emerges necessarily when some pairs of neurons are more likely to be connected than others. Our numerical results imply that such overrepresentation can also be sustained when connection probabilities are only approximately symmetric.

Shape Memory Alloy Driven Two-Way Deformation Structure Design

Shape memory alloy (SMA) is a type of smart material with huge driving force, and large driven displacement. Using SMA as the structure driver may force a bi-directional deformation structure that can achieve high efficiency and compact driving structure. SMA, together with elastic component, forms the two-way driving structure; however the two, if respectively serving as the driver, can enable the occurrence of the simplest bi-directional drivable structure. The offset structure constituted by SMA, not only can achieve two-way drive, and does not require continuous input of energy during the shape-stabilizing period. This paper studies the design methods of these two structures, gives the typical formulas, which makes possible efficient and compact bi-directional drivable structures.

Disentangling attention and awareness: The case of predictive learning

Custers and Aarts (2011) demonstrated that whether predictive relations between two events are stored in memory as unidirectional or bi-directional structures does not depend on awareness, but on attention. Here, the role of attention and top-down processes in producing these effects are investigated more closely.

Modeling the Effect of Flow Structure Selection on Residence Time in an Artificial Canal System: Case Study

A novel modeling approach was used to investigate the residence times of Oyster Cove, an artificial canal system connected to adjacent water bodies by unidirectional and bidirectional flow structures. A field program was carried out to evaluate and quantify the exchange of water through the system of flow structures and to gain an understanding of the mixing dynamics within the artificial canal. Results from the field program were also used to validate a three-dimensional circulation model and a flushing model used to quantify the existing residence time of the canal system. Finally, the model was used to compare several hypothetical design alternatives, to identify the effect on the canal's residence time, by changing the positions of the flow structures and using different combinations of structures. The comparison showed the significant improvements in residence times that could be achieved.

Modeling elastic and thermal properties of 2.5D carbon fiber C/SiC hybrid matrix composites by homogenization method

Advanced carbon fiber hybrid carbon-ceramic matrix composites are realizing their potential in many thermostructural components for aerospace vehicles. This work presents ab-initio predictions of elastic constants and thermal properties for 2.5D carbon fiber reinforced carbon-silicon carbide hybrid matrix composites, by using the homogenization technique. The homogenization technique takes properties of individual components of the composites (fiber and matrix) and characteristics of the geometrical architecture of the preform to perform calculations. Ab-initio modeling of mechanical and thermal properties is very attractive, especially during the material development stage, when larger samples may be prohibitively expensive or impossible to fabricate. Modeling is also useful when bigger samples would be prohibitively expensive or impractical. Thermostructural composites made of 2.5D preforms are easy to manufacture in relation to 3D preforms. Besides, 2.5D preforms are also resistant to thermo cycling and have high resistance to crack propagation in relation to ply stacked composites such as unidirectional (1D) and bidirectional (2D) structures. The calculations were performed by setting an overall carbon fiber volume fraction at 40, 45 and 50 for a 2D stacked composite, and volume fraction in Z-direction of 2, 4 and 6.

Occurrence of bidirectional type III bursts in solar flares

The 13 pairs of type III bursts with the bidirectional drift structures recorded with the spectrograph in the frequency ranges of 230-300 MHz and 625-1500 MHz at the Yunnan Observatory and 2600-3800 MHz at the Beijing National Astronomical Observatories are analyzed in this present article and the outstanding characteristics of these events are obtained. These bursts respectively reveal that the separatrix frequency between the bursts with positive and negative drifts comes between 250 MHz and 3420 MHz, with a gap being between 0.6 MHz and 110 MHz; the duration is 53 ms-1880 ms and the frequency drift rate is between 45 MHz/s and 56000 MHz/s. The drift rate at metric wavelengths is relatively low, only a few decades of MHz while it is comparatively high at microwave wavelengths, reaching 56000 MHz/s. The qualitative explanation of these events is given in this paper.

Biconical structures in two-dimensional anisotropic Heisenberg antiferromagnets

Square lattice Heisenberg and $XY$ antiferromagnets with uniaxial anisotropy are studied in a field along the easy axis. Based on ground-state considerations and Monte Carlo simulations, the role of biconical structures in the transition region between the antiferromagnetic and spin-flop phases is analyzed. In particular, on adding a single-ion anisotropy to the $XXZ$ antiferromagnet, one observes, depending on the sign of that anisotropy, either an intervening biconical phase or a direct transition of first order separating the two phases. In the case of the anisotropic $XY$ model, the degeneracy of the ground state, at a critical field, in antiferromagnetic, spin-flop, and bidirectional structures seems to result, as in the case of the $XXZ$ model, in a narrow disordered phase between the antiferromagnetic and spin-flop phases, dominated by bidirectional fluctuations.

New low complexity bidirectional systolic structures for serial multiplication over the finite field GF(qm)

Two low complexity bidirectional systolic structures for serial multiplication over the finite field GF(q/sup m/) are presented. The first structure is based on generalising the GF(2/sup m/) multiplication algorithm and the semi-systolic multiplication structure for general q, and on a novel cell architecture which is equivalent to two conventional cells used in existing structures and has a new fast mod-q multi operand adder. The second proposed structure is a modified version of the first structure where a twin-pipe is used to double the throughput rate and the hardware utilisation. Owing to the use of bidirectional data flow in both structures, the initial delay of the proposed systolic structures is reduced by 50% when compared to related systolic structures based on unidirectional data flow. The area requirements of the new systolic structures are also reduced considerably when compared to related structures. This is illustrated by a remarkable reduction of more than 45% in the area-time requirements of the proposed structures when compared to existing GF(2/sup m/) structures based on unidirectional data flow.