Channel Rank Research Articles

Effects of pulse shape on pitch sensitivity of cochlear implant users

Contemporary cochlear implants (CIs) use cathodic-leading symmetric biphasic (C-BP) pulses for electrical stimulation. It remains unclear whether asymmetric pulses emphasizing the anodic or cathodic phase may improve spectral and temporal coding with CIs. This study tested place- and temporal-pitch sensitivity with C-BP, anodic-centered triphasic (A-TP), and cathodic-centered triphasic (C-TP) pulse trains on apical, middle, and basal electrodes in 10 implanted ears. Virtual channel ranking (VCR) thresholds (for place-pitch sensitivity) were measured at both a low and a high pulse rate of 99 (Experiment 1) and 1000 (Experiment 2) pulses per second (pps), and amplitude modulation frequency ranking (AMFR) thresholds (for temporal-pitch sensitivity) were measured at a 1000-pps pulse rate in Experiment 3. All stimuli were presented in monopolar mode. Results of all experiments showed that detection thresholds, most comfortable levels (MCLs), VCR thresholds, and AMFR thresholds were higher on more basal electrodes. C-BP pulses had longer active phase duration and thus lower detection thresholds and MCLs than A-TP and C-TP pulses. Compared to C-TP pulses, A-TP pulses had lower detection thresholds at the 99-pps but not the 1000-pps pulse rate, and had lower MCLs at both pulse rates. A-TP pulses led to lower VCR thresholds than C-BP pulses, and in turn than C-TP pulses, at the 1000-pps pulse rate. However, pulse shape did not affect VCR thresholds at the 99-pps pulse rate (possibly due to the fixed temporal pitch) or AMFR thresholds at the 1000-pps pulse rate (where the overall high performance may have reduced the changes with different pulse shapes). Notably, stronger polarity effect on VCR thresholds (or more improvement in VCR with A-TP than with C-TP pulses) at the 1000-pps pulse rate was associated with stronger polarity effect on detection thresholds at the 99-pps pulse rate (consistent with more degeneration of auditory nerve peripheral processes). The results suggest that A-TP pulses may improve place-pitch sensitivity or spectral coding for CI users, especially in situations with peripheral process degeneration.

Simulation of open quantum systems via low-depth convex unitary evolutions

Simulating physical systems on quantum devices is one of the most promising applications of quantum technology. Current quantum approaches to simulating open quantum systems are still practically challenging on NISQ-era devices, because they typically require ancilla qubits and extensive controlled sequences. In this work, we propose a hybrid quantum-classical approach for simulating a class of open system dynamics called random-unitary channels. These channels naturally decompose into a series of convex unitary evolutions, which can then be efficiently sampled and run as independent circuits. The method does not require deep ancilla frameworks and thus can be implemented with lower noise costs. We implement simulations of open quantum systems up to dozens of qubits and with large channel ranks. Published by the American Physical Society 2024

High frequency oscillation network dynamics predict outcome in non-palliative epilepsy surgery.

High frequency oscillations are a promising biomarker of outcome in intractable epilepsy. Prior high frequency oscillation work focused on counting high frequency oscillations on individual channels, and it is still unclear how to translate those results into clinical care. We show that high frequency oscillations arise as network discharges that have valuable properties as predictive biomarkers. Here, we develop a tool to predict patient outcome before surgical resection is performed, based on only prospective information. In addition to determining high frequency oscillation rate on every channel, we performed a correlational analysis to evaluate the functional connectivity of high frequency oscillations in 28 patients with intracranial electrodes. We found that high frequency oscillations were often not solitary events on a single channel, but part of a local network discharge. Eigenvector and outcloseness centrality were used to rank channel importance within the connectivity network, then used to compare patient outcome by comparison with the seizure onset zone or a proportion within the proposed resected channels (critical resection percentage). Combining the knowledge of each patient's seizure onset zone resection plan along with our computed high frequency oscillation network centralities and high frequency oscillation rate, we develop a Naïve Bayes model that predicts outcome (positive predictive value: 100%) better than predicting based upon fully resecting the seizure onset zone (positive predictive value: 71%). Surgical margins had a large effect on outcomes: non-palliative patients in whom most of the seizure onset zone was resected ('definitive surgery', ≥ 80% resected) had predictable outcomes, whereas palliative surgeries (<80% resected) were not predictable. These results suggest that the addition of network properties of high frequency oscillations is more accurate in predicting patient outcome than seizure onset zone alone in patients with most of the seizure onset zone removed and offer great promise for informing clinical decisions in surgery for refractory epilepsy.

Towards performance-maximizing neural network pruning via global channel attention

Network pruning has attracted increasing attention recently for its capability of transferring large-scale neural networks (e.g., CNNs) into resource-constrained devices. Such a transfer is typically achieved by removing redundant network parameters while retaining its generalization performance in a static or dynamic manner. Concretely, static pruning usually maintains a larger and fit-to-all (samples) compressed network by removing the same channels for all samples, which cannot maximally excavate redundancy in the given network. In contrast, dynamic pruning can adaptively remove (more) different channels for different samples and obtain state-of-the-art performance along with a higher compression ratio. However, since the system has to preserve the complete network information for sample-specific pruning, the dynamic pruning methods are usually not memory-efficient. In this paper, our interest is to explore a static alternative, dubbed GlobalPru, from a different perspective by respecting the differences among data. Specifically, a novel channel attention-based learn-to-rank framework is proposed to learn a global ranking of channels with respect to network redundancy. In this method, each sample-wise (local) channel attention is forced to reach an agreement on the global ranking among different data. Hence, all samples can empirically share the same ranking of channels and make the pruning statically in practice. Extensive experiments on ImageNet, SVHN, and CIFAR-10/100 demonstrate that the proposed GlobalPru achieves superior performance than state-of-the-art static and dynamic pruning methods by significant margins.

DTLR-CS: Deep tensor low rank channel cross fusion neural network for reproductive cell segmentation.

In recent years, with the development of deep learning technology, deep neural networks have been widely used in the field of medical image segmentation. U-shaped Network(U-Net) is a segmentation network proposed for medical images based on full-convolution and is gradually becoming the most commonly used segmentation architecture in the medical field. The encoder of U-Net is mainly used to capture the context information in the image, which plays an important role in the performance of the semantic segmentation algorithm. However, it is unstable for U-Net with simple skip connection to perform unstably in global multi-scale modelling, and it is prone to semantic gaps in feature fusion. Inspired by this, in this work, we propose a Deep Tensor Low Rank Channel Cross Fusion Neural Network (DTLR-CS) to replace the simple skip connection in U-Net. To avoid space compression and to solve the high rank problem, we designed a tensor low-ranking module to generate a large number of low-rank tensors containing context features. To reduce semantic differences, we introduced a cross-fusion connection module, which consists of a channel cross-fusion sub-module and a feature connection sub-module. Based on the proposed network, experiments have shown that our network has accurate cell segmentation performance.

A CNN Compression Method via Dynamic Channel Ranking Strategy

In recent years, the rapid development of mobile devices and embedded system raises a demand for intelligent models to address increasingly complicated problems. However, the complexity of the structure and extensive parameters press significantly on efficiency, storage space, and energy consumption. Additionally, the explosive growth of tasks with enormous model structures and parameters makes it impossible to compress models manually. Thus, a standardized and effective model compression solution achieving lightweight neural networks is established as an urgent demand by the industry. Accordingly, Dynamic Channel Ranking Strategy (DCRS) method is proposed to compress deep convolutional neural networks. DCRS selects channels with high contribution of each prunable layer according to compression ratio searched by reinforcement learning agent. Compared with current model compression methods, DCRS efficaciously applies various channel ranking strategies on prunable layers. Experiments indicate with a 50% compression ratio, compressed MobileNet achieved 70.62% top1 and 88.2% top5 accuracy on ImageNet, and compressed ResNet achieved 92.03% accuracy on CIFAR-10. DCRS reduces more FLOPS in these neural networks. The compressed model achieves the best Top-1 and Top-5 accuracy on ResNet50, the best Top-1 accuracy on MobilNetV1.

Micro-expression recognition based on EEG signals

Research into micro-expression recognition is increasingly motivated by its potential applications to mental-health diagnosis and policing. Thus far, micro-expression recognition research and applications have been based on computer vision processing technology, with notable limitations. For instance, changes in ambient lighting or in the head posture and facial occlusion can all affect micro-expression recognition. The high temporal resolution of EEG technology captures brain activity associated with micro-expressions, and hence objectively identifies micro-expressions from a neurophysiological perspective. Herein, we report a real-time supervision and emotional expression suppression (SEES) experimental paradigm. We collected micro-expression video and EEG data from 68 subjects under positive emotion induction. Thereafter, the EEG frequency-band power, statistical features, autoregressive (AR) model parameters, and wavelet entropy were extracted as features. We ranked all 128 channels based on the micro-expression recognition task classification accuracy using all the features extracted for each channel. Based on the channel ranking, different sizes of channel combinations were chosen. Subsequently, the optimal subset of these features from different combinations was selected using a sequential backward selection (SBS) algorithm and fed into different classifiers, including a support vector machine (SVM), K-nearest neighbor (KNN), gradient-boosted decision tree (GBDT), and back-propagation (BP) neuron network. The different classifiers achieve promising accuracy for micro-expression recognition, with GBDT exhibiting the highest accuracy of 85.96 % when 11 channels were selected. This demonstrates the feasibility of recognizing micro-expressions based on EEG signals.

Low-Rank Matrix Sensing-Based Channel Estimation for mmWave and THz Hybrid MIMO Systems

This paper studies the channel estimation for wideband multiple-input multiple-output (MIMO) systems equipped with hybrid analog/digital transceivers operating in the millimeter-wave (mmWave) or terahertz (THz) bands. By exploiting the low-rank property of the concatenated channel matrix of the delay taps, we formulate the channel estimation problem as a low-rank matrix sensing (LRMS) problem and solve it using a low-complexity generalized conditional gradient-alternating minimization (GCG-ALTMIN) algorithm. This LRMS-based solution can accommodate different precoder/combiner and training structures. In addition, it does not require knowledge about the array responses at the transceivers, in contrast to most existing solutions allowing low training overhead. Furthermore, a preconditioned conjugate gradient (PCG) algorithm-based implementation and a low-rank matrix completion (LRMC) formulation are proposed to further reduce the computational complexity. In order to enhance the channel estimation performance for fat and tall channel matrices, we introduce a matrix reshaping approach that can preserve the channel rank by exploiting the shift-invariance property of uniform arrays. We also introduce a spectrum denoising (SD) approach for further improving the performance when the array responses are known and the number of paths is small. These approaches can effectively enhance the performance at a given training overhead. Simulation results suggest that the proposed solutions can achieve higher channel estimation accuracy and reduce the computational complexity as compared to several representative channel estimation schemes.

Simulated Clear-Sky Water Vapor and Temperature Retrievals from PREFIRE Measurements

Abstract The Polar Radiant Energy in the Far Infrared Experiment (PREFIRE) mission will measure Earth’s emission at wavelengths ranging from 3 to 54 μm. The prelaunch clear-sky retrieval algorithm, evaluated with simulated test data, indicates that PREFIRE measurements will be valuable for retrieving atmospheric water vapor and temperature profiles. Far infrared measurements provide unique retrieval information, indicated by the high ranking of select FIR channels as primary contributors to the total degrees of freedom for signal (DFS). In utilizing all the PREFIRE channels, the average total DFS of 4 test regions ranges from 1.90 to 4.71. The information content increases with higher column water vapor and in the presence of near-surface temperature inversions. Using the DFS profiles for guidance, the retrieval concentrates information into 7 distinct layers to reduce the retrieval uncertainty per layer. Sensitivity tests indicate forward model error due to surface emissivity uncertainty results in about a 9% increase in column water vapor uncertainty. The clear-sky retrieval is sensitive to the presence of undetected ice clouds, especially those with optical depths larger than 0.2. Hence, in addition to a separate PREFIRE cloud mask, optimal estimation retrieval metrics are explored as possible indicators of cloudy scenes.

Quantum Dynamics is Not Strictly Bidivisible.

We address the question of the existence of quantum channels that are divisible in two quantum channels but not in three or, more generally, channels divisible in n but not in n+1 parts. We show that for the qubit those channels do not exist, whereas for general finite-dimensional quantum channels the same holds at least for full Kraus rank channels. To prove these results, we introduce a novel decomposition of quantum channels which separates them into a boundary and Markovian part, and it holds for any finite dimension. Additionally, the introduced decomposition amounts to the well-known connection between divisibility classes and implementation types of quantum dynamical maps and can be used to implement quantum channels using smaller quantum registers.

Assessing the Relationship Between Pitch Perception and Neural Health in Cochlear Implant Users.

Various neural health estimates have been shown to indicate the density of spiral ganglion neurons in animal and modeling studies of cochlear implants (CIs). However, when applied to human CI users, these neural health estimates based on psychophysical and electrophysiological measures are not consistently correlated with each other or with the speech recognition performance. This study investigated whether the neural health estimates have stronger correlations with the temporal and place pitch sensitivity than with the speech recognition performance. On five electrodes in 12 tested ears of eight adult CI users, polarity effect (PE), multipulse integration (MPI), and interphase gap (IPG) effect on the amplitude growth function (AGF) of electrically evoked compound action potential (ECAP) were measured to estimate neural health, while thresholds of amplitude modulation frequency ranking (AMFR) and virtual channel ranking (VCR) were measured to indicate temporal and place pitch sensitivity. AzBio sentence recognition in noise was measured using the clinical CI processor for each ear. The results showed significantly poorer AMFR and VCR thresholds on the basal electrodes than on the apical and middle electrodes. Across ears and electrodes, only the IPG offset effect on ECAP AGF had a nearly significant negative correlation with the VCR threshold after removing the outliers. No significant across-ear correlations were found between the mean neural health estimates, mean pitch-ranking thresholds, and AzBio sentence recognition score. This study suggests that the central axon demyelination reflected by the IPG offset effect may be important for the place pitch sensitivity of CI users and that the IPG offset effect may be used to predict the perceptual resolution of virtual channels for CI programming.

Tensor Rank Regularization with Bias Compensation for Millimeter Wave Channel Estimation

This paper presents a novel method of tensor rank regularization with bias compensation for channel estimation in a hybrid millimeter wave MIMO-OFDM system. Channel estimation is challenging due to the unknown number of multipath components that determines the channel rank. In general, finding the intrinsic rank of a tensor is a non-deterministic polynomial-time (NP) hard problem. However, by leveraging the sparse characteristics of millimeter wave channels, we propose a modified CANDECOMP/PARAFAC (CP) decomposition-based method that jointly estimates the tensor rank and channel component matrices. Our approach differs from most existing works that assume the number of channel paths is known and the proposed method is able to estimate channel parameters accurately without the prior knowledge of number of multipaths. The objective of this work is to estimate the tensor rank by a novel sparsity-promoting prior that is incorporated into a standard alternating least squares (ALS) function. We introduce a weighting parameter to control the impact of the previous estimate and the tensor rank estimation bias compensation in the regularized ALS. The channel information is then extracted from the estimated component matrices. Simulation results show that the proposed scheme outperforms the baseline l1 strategy in terms of accuracy and robustness. It also shows that this method significantly improves rank estimation success at the expense of slightly more iterations.

Analysis of Big Data in Social Network Marketing and Social Media

Data is the foundation of every system. Data keep scaling at different types. The creation of an automated system in the field of digital media deals with the enormous amounts of data that accumulate over time on www. Social networks were used to update it globally every second. Social networks offer the chance to engage with others in a personal, professional, and entertaining way. This study tries to show how social networking and media recommendation use a tremendous quantity of data. There are many different social networking websites on www. The primary goal of this study is to examine how individuals interact with social media and how TV media use social media to rank channels and shows. Big data, big data analytics, social networks analytics, TV media social networks, emotional analysis.Every economy, production, organization, business function, and individual now depend on data in some way. The use of the internet, smartphones, social networks, the development of ubiquitous computing, and several other technical breakthroughs are all contributing to the world's expanding data volume. Cost savings, quicker, better decision-making, and the development of new goods and services are all benefits of big data. However, the three V's of big data—Velocity, Volume, and Variety—amplify security and privacy vulnerabilities. These variables include things like massive cloud infrastructures, variety in data sources and formats, acquisition of data occurring in streams, and a rising amount of intercloud migrations. We have examined how the Bloom filter (BF) can be used to overcome problems with sorting and decomposing huge amounts of data both in space and time.

Spectral classification of space agricultural landscape images using ranking of channels according to the highest informativity

This paper explores the question of classification of a series of space images for analysis of the processes of growing crops, using the spectral channels of minimum dimension. This study attempts to give a partial answer to the question: how to detect the problems in growing processes much faster and with a high level of accuracy? In this paper, we report on the results of the survey conducted to determine the existing methods and tools for spectral analysis of multispectral space images for solving forecasting and search problems of mapping the agricultural objects. The present work extends the use of the general image classification process by the computational procedure of ranking spectral channels according to the value of spectral information. The structure of the classification rule based on the informativity indicator of the spectral channel has been developed. The proposed prototype of a geospatial recognition system will provide a digital stratification landscape plan.

Clustered and Distributed Caching Methods for F-RAN-Based mmWave Communications

Fog-radio access networks (F-RANs) alleviate fronthaul delays for cellular networks as compared to their cloud counterparts. This allows them to be suitable solutions for networks that demand low propagation delays. Namely, they are suitable for millimeter wave (mmWave) operations that suffer from short propagation distances and possess a poor scattering environment (low channel ranks). The F-RAN here is comprised of fog nodes that are collocated with radio remote heads (RRHs) to provide local processing capabilities for mobile station (MS) terminals. These terminals demand various network functions (NFs) that correspond to different service requests. Now, provisioning these NFs on the fog nodes also yields service delays due to the requirement for service migration from the cloud, i.e., offloading to the fog nodes. One solution to reduce this service delay is to provide cached copies of popular NFs in advance. Hence, it is critical to study function popularity and allow for content caching at the F-RAN. This is further a necessity given the limited resources at the fog nodes, thus requiring efficient resource management to enhance network capacity at reduced power and cost penalty. This paper proposes novel solutions that allocate popular NFs on the fog nodes to accelerate services for the terminals, namely, the clustered and distributed caching methods. The two methods are analyzed and compared against the baseline uncached provisioning schemes in terms of service delay, energy consumption, and cost.

FMCPR: Flexible Multiparameter-Based Channel Prediction and Ranking for CR-Enabled Massive IoT

The cognitive radio-enabled <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$massive$ </tex-math></inline-formula> Internet of Things (CR- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$m$ </tex-math></inline-formula> IoT) is envisioned to shape the future of densely connected IoT devices in the sixth-generation networks to support the hyperconnected society. In conventional CR networks, secondary users (SUs) sense the whole block of spectrum to find idle channels, which is an energy-consuming, delay-inducing, and processing-intensive task. With the large scale of resource-constrained heterogeneous devices in CR- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$m$ </tex-math></inline-formula> IoT, the sensing process becomes a major hurdle for CR- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$m$ </tex-math></inline-formula> IoT devices to achieve efficient utilization of the limited device and network resources. Thus, a novel multiparameter-based flexible scheme is proposed for idle channel prediction and channel ranking, which considers priorities as well as heterogeneity of users. The scheme uses a probabilistic approach and employs multiple parameters simultaneously to evaluate the suitability of a channel before selecting it for transmission. In addition, valid channel obsolescence, a major problem inherent with channel prediction and ranking, is countered by the proposed scheme. The scheme is evaluated under the impact of variable primary and SUs’ arrivals and under multiple channel failures rates and variable sensing and frame time duration. The proposed scheme is also compared with its own modified version that disregards channel failures, and with the random channel selection approach followed by IEEE 802.22. The overall evaluation is conducted under realistic spectrum sensing. Simulation results show that for different parameter values, the proposed scheme improves the collision probability by 11%–55%, reduces sensing time and energy by 60% and 65%, respectively, and enhances throughput by 4%–70%, and spectrum utilization efficiency by 11%–40%.

On the Fundamental Characteristics of Intelligent Reflecting Surface Enabled MIMO Channels

Smart radio environment (SRE) via reconfigurable and intelligent reflecting surfaces (IRSs) is envisioned to shape wireless communication beyond 5G. In SRE, low-cost, low-power, and passive IRS elements will, in an intelligent manner, alter the phase shifts of incident waves to beamform the reflected signal in the desired direction. SRE challenges the status quo of communication system design as the wireless channel is no longer an impairment but a customizable component for enhancing system performance. In this article, we systematically investigate and visualize the performance of IRS-multiple-input multiple-output (MIMO) channels in a practical outdoor application scenario. Here we use channel gain, rank, condition number, and spectral efficiency heat maps of the coverage area to reveal several underlying characteristics of IRS-MIMO channels under different propagation conditions and array configurations. We also present several observations and insights that have the potential to contribute design guidelines toward IRS-assisted MIMO communication for next-generation networks.

All-Optical Distributed MIMO for LiFi: Spatial Diversity Versus Spatial Multiplexing

LiFi or networked optical wireless communication is likely to play an important role in offloading mobile data traffic from radio into the optical spectrum. As the number of Internet of Things (IoT) devices is growing, the RF spectrum becomes a rare resource. Imaging IoT sensors like cameras, ultrasonic devices, and Lidars have real-time requirements, need a high-capacity uplink, and operate in environments that cause or are sensitive against electromagnetic interference. In this paper, for the first time, we present realtime communication over an all-optical fixed-wireless LiFi link based on the distributed multiple-input multiple-output concept. For distributing the wireless signals, plastic optical fibers are used as an analog front-haul. We study the operation of the distributed multiple-input multiple-output link in two modes, i.e., spatial diversity and spatial multiplexing. For the diversity mode, a new combiner is presented, which can support equal gain as well as selection combining. We demonstrate that selection combining is highly effective and enables a similar LiFi performance in up- and downlink, as it is desirable for industrial applications. For the spatial multiplexing mode, we observe that the channel rank and the achievable throughput depend strongly on the user location. As effective solutions, we study the benefits of angular diversity and multiple-input multiple-output mode switching together with multi-user multiplexing and conclude that a dynamic switching between spatial diversity and spatial multiplexing is a practical approach.

Rethinking the Ranks of Visual Channels.

Data can be visually represented using visual channels like position, length or luminance. An existing ranking of these visual channels is based on how accurately participants could report the ratio between two depicted values. There is an assumption that this ranking should hold for different tasks and for different numbers of marks. However, there is surprisingly little existing work that tests this assumption, especially given that visually computing ratios is relatively unimportant in real-world visualizations, compared to seeing, remembering, and comparing trends and motifs, across displays that almost universally depict more than two values. To simulate the information extracted from a glance at a visualization, we instead asked participants to immediately reproduce a set of values from memory after they were shown the visualization. These values could be shown in a bar graph (position (bar)), line graph (position (line)), heat map (luminance), bubble chart (area), misaligned bar graph (length), or 'wind map' (angle). With a Bayesian multilevel modeling approach, we show how the rank positions of visual channels shift across different numbers of marks (2, 4 or 8) and for bias, precision, and error measures. The ranking did not hold, even for reproductions of only 2 marks, and the new probabilistic ranking was highly inconsistent for reproductions of different numbers of marks. Other factors besides channel choice had an order of magnitude more influence on performance, such as the number of values in the series (e.g., more marks led to larger errors), or the value of each mark (e.g., small values were systematically overestimated). Every visual channel was worse for displays with 8 marks than 4, consistent with established limits on visual memory. These results point to the need for a body of empirical studies that move beyond two-value ratio judgments as a baseline for reliably ranking the quality of a visual channel, including testing new tasks (detection of trends or motifs), timescales (immediate computation, or later comparison), and the number of values (from a handful, to thousands).

DLSTM-Based Successive Cancellation Flipping Decoder for Short Polar Codes.

Polar code has been adopted as the control channel coding scheme for the fifth generation (5G), and the performance of short polar codes is receiving intensive attention. The successive cancellation flipping (SC flipping) algorithm suffers a significant performance loss in short block lengths. To address this issue, we propose a double long short-term memory (DLSTM) neural network to locate the first error bit. To enhance the prediction accuracy of the DLSTM network, all frozen bits are clipped in the output layer. Then, Gaussian approximation is applied to measure the channel reliability and rank the flipping set to choose the least reliable position for multi-bit flipping. To be robust under different codewords, padding and masking strategies aid the network architecture to be compatible with multiple block lengths. Numerical results indicate that the error-correction performance of the proposed algorithm is competitive with that of the CA-SCL algorithm. It has better performance than the machine learning-based multi-bit flipping SC (ML-MSCF) decoder and the dynamic SC flipping (DSCF) decoder for short polar codes.