Vicente Botella-Soler (173290), Stéphane Deny (5204018), Georg Martius (415557) et al. · Figshare · 2018
<p><b>A</b>: Shuffles of responses to repeated stimulus presentations remove different types of correlations, but preserve average locking to the stimulus (PSTH), and thus stimulus-induced correlations. <b>B</b>: A repeated stimulus fragment (red trace), nonlinear kernelized decoder predictions using real responses (green), and using responses without different types of correlations (gray); shown is the prediction mean ± SD over repeats. <b>C</b>: Increase in decoding error (MSE) when spike-history dependencies or noise correlations are removed (average ± SEM across sites); percentages report fractional differences relative to the original performance. <b>D</b>: Spike count distributions for a single example cell. Removing spike-history dependencies broadens the distributions, in particular in constant epochs. Dashed line = expectation for a fully randomized spike train with a matched firing rate. <b>E</b>: Variance-to-mean ratio <i>F</i> of spike count distributions for spike trains with and without spike-history dependencies. Each point is a cell that contributes most to decoding at a particular site (when the same cell contributes to multiple sites, average ± SD across sites is shown).</p>
Cited by 0Open access
Roman Kessler, Alexander Enge, Michael A. Skeide · Communications Biology · 2025
Cited by 10Open access
Anuranjan Sethi · SSRN Electronic Journal · 2023
Cited by 0Open access
Roman Kessler, Alexander Enge, Michael A. Skeide · arXiv (Cornell University) · 2024
EEG preprocessing varies widely between studies, but its impact on classification performance remains poorly understood. To address this gap, we analyzed seven experiments with 40 participants drawn from the public ERP CORE dataset. We systematically varied key preprocessing steps, such as filtering, referencing, baseline interval, detrending, and multiple artifact correction steps. Then we performed trial-wise binary classification (i.e., decoding) using neural networks (EEGNet), or time-resolved logistic regressions. Our findings demonstrate that preprocessing choices influenced decoding performance considerably. All artifact correction steps reduced decoding performance across experiments and models, while higher high-pass filter cutoffs consistently increased decoding performance. For EEGNet, baseline correction further increased decoding performance, and for time-resolved classifiers, linear detrending, and lower low-pass filter cutoffs increased decoding performance. The influence of other preprocessing choices was specific for each experiment or event-related potential component. The current results underline the importance of carefully selecting preprocessing steps for EEG-based decoding. While uncorrected artifacts may increase decoding performance, this comes at the expense of interpretability and model validity, as the model may exploit structured noise rather than the neural signal.
Cited by 2Open access
Lubomir Kostal, Petr Lánský, Stevan Pilarski · Journal of Neural Engineering · 2015
We demonstrate the existence of such threshold regions in early auditory and olfactory information coding. We discuss the origin of the threshold effect and its impact on the design of effective coding approaches in terms of relevant population size.
Cited by 10
Stefan K. Ehrlich (6475490), Emmanuel Dean-Leon (16528239), Nicholas Tacca (16528242) et al. · Figshare · 2023
<p>Average across subjects (n = 11) single-trial classification performance (M±SD [%]) for decoding anticipation of variants of takeover (<i>HR</i>, <i>RH</i>) against non-takeover situations (<i>HH</i>, <i>RR</i>) for sequential collaboration (<i>sC</i>) and intermittent collaboration (<i>iC</i>). Bolded numbers show results that are significantly above the sample-size adjusted chance-level <i>p</i><sub><i>chance</i></sub> = 54.3%. Detailed per subject results of decoding performance are reported in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0287958#pone.0287958.s003" target="_blank">S3 Appendix</a>.</p>
Cited by 0Open access
Arthur Mensch (10739725), Julien Mairal (10739728), Bertrand Thirion (40674) et al. · Figshare · 2021
<p><b>(A)</b> Multi-study decoding improves the performance of cognitive task prediction across subjects for most studies. <b>(B)</b> Overall, decoding from task-optimized networks leads to a mean improvement accuracy of 5.8% compared to voxel or networks based approaches. Each point corresponds to a study and a train/test split. <b>(C)</b> Studies of typical size strongly benefit from transfer learning, whereas little information is gained for very large studies. <b>(D)</b> Contrasts that are moderately difficult to decode benefit most from transfer. Error bars are calculated over 20 random data half-split. *<b>(D)</b> shows <i>per-contrast</i> balanced accuracy (50% chance level), whereas <i>per-study</i> classification accuracy is used everywhere else. Numbers are reported in Table A in <a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1008795#pcbi.1008795.s001" target="_blank">S1 Appendix</a>.</p>
Cited by 0Open access
R. G. Marquart · IRE Transactions on Communications Systems · 1967
Forced-erasure decoding, an easily implementable technique for realizing the fullest capability of a code, is described and evaluated. This type of decoding combines the implementation simplicity of digital decoding with the performance enhancement expected of correlation decoding techniques. Adaptive versions, with and without interleaving for fading channels, are also treated. The simulation of representative channels is described, and the performances of forced-erasure decoding, correlation decoding, and digital decoding on these channels are determined and compared.
Cited by 3
Jean Michel de Souza Sant’Ana, Arliones Hoeller, Richard Demo Souza et al. · University of Oulu Repository (University of Oulu) · 2020
This letter considers the use of successive interference cancellation (SIC) to decode superposed signals in Long Range (LoRa) networks. We build over a known stochastic geometry model for LoRa networks and include the effect of recovering colliding packets through SIC. We derive closed-form expressions for the successful decoding of packets using SIC taking path loss, fading, noise, and interference into account, while we validate the model by means of Monte Carlo simulations. Results show that SIC-enabled LoRa networks improve worst-case reliability by up to 34%. We show that, for at least one test scenario, SIC increases by 159% the number of served users with the same worst-case reliability level.
Cited by 27
Amirhossein Pourabdollah · Open Science Framework · 2025
This project constitutes the computational preprocessing phase of a Master's thesis investigating the Neurovisceral Integration Model. It implements a reproducible signal processing pipeline to transform raw Electrocardiography (ECSMP dataset) signals into validated Heart Rate Variability (HRV) biomarkers. By isolating specific autonomic indices (Time-domain, Frequency-domain, and Non-linear), this repository generates the physiological feature matrix required to empirically test the relationship between autonomic regulation and executive function.
Cited by 0Open access
Soljanin, E., Urbanke, R. · Infoscience (Ecole Polytechnique Fédérale de Lausanne) · 1997
Cited by 0Open access
Shahrbaf Motlagh, Meysam · UWSpace (University of Waterloo) · 2021
In a massive MIMO system, base stations (BS) utilize a large number of antennas to simultaneously serve several (single or multi-antenna) users at once, where the number of BS antennas is normally assumed to be significantly larger than the number of users. In massive MIMO systems operating in time division duplex (TDD) mode, the channel state information (CSI) is estimated via uplink pilot sequences that are orthogonal in a cell but re-used in other cells. Re-using the pilots, however, contaminates the CSI estimate at BSs by the channel of the users sharing the same pilot in other cells; thus causing pilot contamination which creates coherent interference that, as the number of BS antennas grows, scales at the same rate as the desired signal. Hence, in the asymptotic limits of large antennas, the effects of non-coherent interference terms and noise disappear, except for the pilot contamination interference. A common technique used in the literature to deal with this interference is to treat it as noise (TIN). When using TIN, users' throughput will converge to a constant and thus the benefits of using an ever greater number of BS antennas saturate. However, it is known that the use of TIN in interference networks is only preferred in the weak interference regime, and it is sub-optimal in other regimes (e.g., moderate or strong interference). In this thesis, we show that as the number of BS antennas increases, the pilot contamination interference is no longer weak, and therefore it is beneficial to treat it differently (e.g., decode it jointly with the desired signal) to improve users’ throughput. \n \n \nIn the first part of the thesis, we study the performance of interference decoding schemes based on simultaneous unique decoding (SD) and simultaneous non-unique decoding (SND), and show that by doing so the rate saturation effect is eliminated as the number of antennas increases; hence, the per-user rates grow unbounded. We analytically study the performance of two well-known linear combining/precoding methods, namely, MRC/MRT and ZF, for spatially correlated/uncorrelated Rayleigh fading channel models, and obtain closed-form expressions of rate lower bounds for these using a worst-case uncorrelated noise technique for multi-user channels. We compare the performance of the different interference management schemes, TIN/SD/SND, based on the maximum symmetric rate they can offer to the users. Specifically, we first obtain structural results for a symmetric two-cell setting as well as the high SINR regime, that provide insights into the benefits of using interference decoding schemes in different regimes of number of BS antennas. We numerically illustrate the performance of the different schemes and show that with a practical number of antennas, SND strictly outperforms TIN. This gain improves with increasing the number of antennas, and also ZF performs significantly better than MRC/MRT due to better mitigation of multi-user interference. Furthermore, we study the performance of regularized ZF (RZF) via Monte Carlo simulations, and observe that it achieves better rates than ZF for moderately small number of antennas only. Lastly, we numerically investigate the impact of increasing the number of cells, the cell radius, the number of users, the correlation of the channel across antennas and the degree of shadow fading on system performance. \n \n \nIn the second part of the thesis, we study the performance of partial interference decoding based on rate splitting (RS) and non-unique decoding. Specifically, we propose to partition each user’s message into two independent layers, and partially decode the pilot contamination interference while treating the remaining part as noise based on a power splitting strategy. In particular, for a two-cell system, we investigate the benefits of an RS scheme based on the celebrated Han-Kobayashi (HK) region, which provides the best known achievable performance for a two-user interference channel (IC). In the case of more than two cells, we propose a generalized RS scheme that non-uniquely decodes each layer of the pilot contamination interference and uses only one power splitting coefficient per IC. In addition, we establish an achievable region for this generalized RS scheme using the non-unique decoding technique. In both cases of two cells and more than two cells and for a practical number of antennas, we numerically study the performance of the proposed RS schemes by numerically optimizing the power splitting coefficients, and show that they achieve significantly higher rates than TIN/SD/SND in all scenarios. Similar to the first part of the thesis, we also numerically examine the impact of increasing the number of cells, the cell radius, the number of users, the correlation of the channel across antennas and the degree of shadow fading on the performance of the RS schemes. Lastly, our simulation results reveal that by replacing the numerically optimized values of the power splitting coefficients with their pre-computed average values (over a large number of realizations), the performance loss is quite negligible, thus reducing the optimization complexity.
Cited by 0Open access
Moemi Matsuo · Open access Journal of Neurology & Neurosurgery · 2024
Background: Electroencephalogram patterns help in evaluating the extent of ischemic brain injury and predicting functional performance. Aim: To determine a possible correlation between attentional task performance and electroencephalogram waves. Methods: The cerebral activity of 12 healthy young adults was investigated using an electroencephalogram while they underwent the TrailMaking Test-A and B as attentional tasks. Results: A significant correlation was observed between a stronger occipital delta power during rest and higher error rates, as well as weaker temporal and central delta power during the Trail-Making Test-B and longer task completion times. Delta waves during both the resting-state and task conditions correlated with task performance, which might be affected by the induced cerebral lobes. Conclusion: The default mode network might predict attention deficits. Our findings further our understanding of the correlation between the default mode network and attentional task performance
Cited by 2Open access
Anonymous · Physical Review Applied · 2026
Cited by 0
Sankariah, S., Eswaran, Chikkannan · Siti Hasmah Digital Library-MMU Institutiona Repository (Multimedia University) · 2018
Cited by 0
Kohtaro Watanabe, Ryusei Kaguchi, Toshiya Shinoda · EURASIP Journal on Wireless Communications and Networking · 2021
Cited by 4Open access
Evangelos Papagiannis, Marcel Ambroze, M. Tomlinson · 2004
The paper presents a method to significantly improve the convergence of iteratively decoded concatenated schemes and reduce the gap between iterative and maximum likelihood (ML) decoding. It is shown that many of the error blocks produced by the iterative decoder can be corrected by modifying a single critical coordinate (channel value) of the received vector and repeating the decoding. This is the basis of the RVCM (received vector coordinate modification) algorithm. Its description, performance and drawbacks are discussed later on. The paper also presents a practically obtained lower bound on ML performance based on the Euclidean distances of the transmitted and the iteratively decoded codewords from the received vector. At low SNR this bound is assuming an unrealistic perfect code, while at high SNR the approximations are getting closer to the real characteristics of the code and the RVCM iterative decoder is shown to achieve the ultimate ML performance.
Cited by 7
Uluğ Bayazıt, William A. Pearlman · Proceedings - International Conference on Image Processing · 2002
The optimal joint decoder utilizing NLIVQ (nonlinear interpolative vector quantization) introduced by Gersho [1990] results in vector quantizers which have reduced encoding complexity at the expense of coding performance loss due to the inferiority of their space-filling property. We show a method of improving a high resolution NLIVQ codebook by partitioning its cells in such a way that the resulting lower resolution codebook consists of cells with better space-filling properties. The resolution reduction method is also extended to the case where the quantizer indices are entropy-constrained. From the simulations it is seen that the unconstrained and constrained entropy versions of the proposed vector quantizer have comparable performance to vector quantizers designed by LEG and ECVQ algorithms.
Cited by 0
Vlsi Implementation, Angelos Spanos, Vassilis Paliouras · 2010
An extension to the HDA tennination criterion, the hardware implementation, and the perfonnance of four stopping criteria for turbo decoders are presented in this paper. The proposed criterion is shown to substantially increase the error correction capabilities in tenns of BER. A perfonnance comparison in tenns of BER of Turbo decoders with fixed number of iterations and early stopping via tennination criteria is presented. It is shown that close to optimal error correcting perfonnance is achieved with minimal hardware increase that can significantly improve decoding throughput.
Cited by 0
IEEE Transactions on Communications · 2004
Performance Evaluation of List Sequence MAP Decoding List sequence (LS) decoding has the potential to yield significant coding gain additional to that of conventional single sequence decoding, and it can be implemented with full backward compatibility in systems where an error-detecting code is concatenated with an error-correcting code. LS maximum-likelihood (ML) decoding provides a list of estimated sequences in likelihood order. For convolutional codes, this list can be obtained with the serial list Viterbi algorithm (SLVA). Through modification of the metric increments of the SLVA, an LS maximum a posteriori (MAP) probability decoding algorithm is obtained that takes into account bitwise a priori probabilities, and produces an ordered list of sequence MAP estimates. The performance of the resulting LS-MAP decoding algorithm is studied in this paper. Computer simulations and approximate analytical expressions, based on geometrical considerations of the decisions domains of LS decoders, are presented. We focus on the frame-error performance of LS-MAP decoding, with genie-assisted error detection, on the additive white Gaussian noise channel. It is concluded that LS-MAP decoding exploits a priori information more efficiently, in order to achieve performance improvements, than does conventional single-sequence MAP decoding. Interestingly, LS-MAP decoding can provide significant improvements at low signal-to-noise ratios, compared with LS-ML decoding. In this environment, it is furthermore observed that feedback convolutional codes offer performance improvements over their feedforward counterparts. Since LS-MAP decoding can be implemented in existing systems at a modest complexity increase, it should have a wide area of applications, such as joint source-channel decoding and other kinds of iterative decoding.
Cited by 0