Q estimation based on the logarithmic spectral area double difference

The Q factor is an essential parameter describing the characteristics of medium absorption within a material during wave propagation. When a seismic wave propagates within the attenuating media, its amplitude decreases and frequency band narrows, resulting in a variation in its logarithmic spectral area. Based on these effects, we calculate the logarithmic spectral area difference (LSAD) before and after attenuation and set a division point to divide the LSAD into two parts. We then compute the difference between the two LSADs to derive a new Q-estimation formula based on computation of the logarithmic spectral area double difference (LSADD). To improve the noise robustness of the Q estimation, we select multiple different division points to calculate the Q factors and consider their average value as our final estimate. We then compare and analyze the noise robustness and bandwidth sensitivity of our technique with other commonly used methods. These results demonstrate that our approach is the most accurate and robust, and least sensitive to the frequency band when processing noisy synthetic seismograms. Finally, we apply our methodology to field vertical seismic profile (VSP) and seismic reflection data, further illustrating the effectiveness of this method to estimate the Q factor.

Noise-trained deep neural networks effectively predict human vision and its neural responses to challenging images

Deep neural networks (DNNs) for object classification have been argued to provide the most promising model of the visual system, accompanied by claims that they have attained or even surpassed human-level performance. Here, we evaluated whether DNNs provide a viable model of human vision when tested with challenging noisy images of objects, sometimes presented at the very limits of visibility. We show that popular state-of-the-art DNNs perform in a qualitatively different manner than humans—they are unusually susceptible to spatially uncorrelated white noise and less impaired by spatially correlated noise. We implemented a noise training procedure to determine whether noise-trained DNNs exhibit more robust responses that better match human behavioral and neural performance. We found that noise-trained DNNs provide a better qualitative match to human performance; moreover, they reliably predict human recognition thresholds on an image-by-image basis. Functional neuroimaging revealed that noise-trained DNNs provide a better correspondence to the pattern-specific neural representations found in both early visual areas and high-level object areas. A layer-specific analysis of the DNNs indicated that noise training led to broad-ranging modifications throughout the network, with greater benefits of noise robustness accruing in progressively higher layers. Our findings demonstrate that noise-trained DNNs provide a viable model to account for human behavioral and neural responses to objects in challenging noisy viewing conditions. Further, they suggest that robustness to noise may be acquired through a process of visual learning.

Structure transforming for constructing constraint force field in musculoskeletal robot

Purpose Human-like musculoskeletal robots can fulfill flexible movement and manipulation with the help of multi joints and actuators. However, in general, sophisticated structures, accurate sensors and well-designed control are all necessary for a musculoskeletal robot to achieve high-precision movement. How to realize the reliable and accurate movement of the robot under the condition of limited sensing and control accuracy is still a bottleneck problem. This paper aims to improve the movement performance of musculoskeletal system by bio-inspired method. Design/methodology/approach Inspired by two kinds of natural constraints, the convergent force field found in neuroscience and attractive region in the environment found in information science, the authors proposed a structure transforming optimization algorithm for constructing constraint force field in musculoskeletal robots. Due to the characteristics of rigid-flexible coupling and variable structures, a constraint force field can be constructed in the task space of the musculoskeletal robot by optimizing the arrangement of muscles. Findings With the help of the constraint force field, the robot can complete precise and robust movement with constant control signals, which brings in the possibility to reduce the requirement of sensing feedback during the motion control of the robot. Experiments are conducted on a musculoskeletal model to evaluate the performance of the proposed method in movement accuracy, noise robustness and structure sensitivity. Originality/value A novel concept, constraint force field, is proposed to realize high-precision movements of musculoskeletal robots. It provides a new theoretical basis for improving the performance of robotic manipulation such as assembly and grasping under the condition that the accuracy of control and sensory are limited.

Noise Robustness Low-Rank Learning Algorithm for Electroencephalogram Signal Classification

Electroencephalogram (EEG) is often used in clinical epilepsy treatment to monitor electrical signal changes in the brain of patients with epilepsy. With the development of signal processing and artificial intelligence technology, artificial intelligence classification method plays an important role in the automatic recognition of epilepsy EEG signals. However, traditional classifiers are easily affected by impurities and noise in epileptic EEG signals. To solve this problem, this paper develops a noise robustness low-rank learning (NRLRL) algorithm for EEG signal classification. NRLRL establishes a low-rank subspace to connect the original data space and label space. Making full use of supervision information, it considers the local information preservation of samples to ensure the low-rank representation of within-class compactness and between-classes dispersion. The asymmetric least squares support vector machine (aLS-SVM) is embedded into the objective function of NRLRL. The aLS-SVM finds the maximum quantile distance between the two classes of samples based on the pinball loss function, which further improves the noise robustness of the model. Several classification experiments with different noise intensity are designed on the Bonn data set, and the experiment results verify the effectiveness of the NRLRL algorithm.

Heterosis Related to Fitness and Phenotypic Variance in the Evolution of Diploid Gene Regulatory Network

Heterosis describes the phenomenon whereby a hybrid population has higher fitness than an inbred population, and has previously been explained by either Mendelian dominance or overdominance, where it is generally assumed that one gene controls one trait. However, recent studies have demonstrated that genes interact through a complex gene regulatory network (GRN). Furthermore, phenotypic variance due to noise is reportedly lower for heterozygotes, whereas the origin of such variance-related heterosis remains elusive. Therefore, a theoretical analysis linking heterosis to GRN evolution and stochastic gene expression dynamics is required. Here, we investigate heterosis related to fitness and phenotypic variance in a system with interacting genes, by numerically evolving diploid GRNs. According to the results, the heterozygote population exhibited higher fitness than the homozygote population, that is, fitness-related heterosis resulting from evolution. In addition, the heterozygote population expressed lower noise-related phenotypic variance in expression levels than the homozygous population, implying that the heterozygote population is more robust to noise. Furthermore, the distribution of the ratio of heterozygote phenotypic variance to homozygote phenotypic variance exhibited quantitative agreement with previous experimental results. By applying dominance and overdominance to the gene expression pattern rather than only a single gene expression, we confirmed the correlation between heterosis and overdominance. We explain our results by proposing that the convex high-fitness region is evolutionarily shaped in the genetic space to gain noise robustness under genetic mixing through sexual reproduction.

Voxel-Wise Adversarial FiboNet for 3D Cerebrovascular Segmentation on Magnetic Resonance Angiography Images

Cerebrovascular segmentation is important in various clinical applications, such as surgical planning and computer-aided diagnosis. In order to achieve high segmentation performance, three challenging problems should be taken into consideration: (1) large variations in vascular anatomies and voxel intensities; (2) severe class imbalance between foreground and background voxels; (3) image noise with different magnitudes. Limited accuracy was achieved without considering these challenges in deep learning-based methods for cerebrovascular segmentation. To overcome the limitations, we propose an end-to-end adversarial model called FiboNet-VANGAN. Specifically, our contributions can be summarized as follows: (1) to relieve the first problem mentioned above, a discriminator is proposed to regularize for voxel-wise distribution consistency between the segmentation results and the ground truth; (2) to mitigate the problem of class imbalance, we propose to use the addition of cross-entropy and Dice coefficient as the loss function of the generator. Focal loss is utilized as the loss function of the discriminator; (3) a new feature connection is proposed, based on which a generator called FiboNet is built. By incorporating Dice coefficient in the training of FiboNet, noise robustness can be improved by a large margin. We evaluate our method on a healthy magnetic resonance angiography (MRA) dataset to validate its effectiveness. A brain atrophy MRA dataset is also collected to test the performance of each method on abnormal cases. Results show that the three problems in cerebrovascular segmentation mentioned above can be alleviated and high segmentation accuracy can be achieved on both datasets using our method.

Scalable and effective multi-level entangled photon states: a promising tool to boost quantum technologies

Multi-level (qudit) entangled photon states are a key resource for both fundamental physics and advanced applied science, as they can significantly boost the capabilities of novel technologies such as quantum communications, cryptography, sensing, metrology, and computing. The benefits of using photons for advanced applications draw on their unique properties: photons can propagate over long distances while preserving state coherence, and they possess multiple degrees of freedom (such as time and frequency) that allow scalable access to higher dimensional state encoding, all while maintaining low platform footprint and complexity. In the context of out-of-lab use, photon generation and processing through integrated devices and off-the-shelf components are in high demand. Similarly, multi-level entanglement detection must be experimentally practical, i.e., ideally requiring feasible single-qudit projections and high noise tolerance. Here, we focus on multi-level optical Bell and cluster states as a critical resource for quantum technologies, as well as on universal witness operators for their feasible detection and entanglement characterization. Time- and frequency-entangled states are the main platform considered in this context. We review a promising approach for the scalable, cost-effective generation and processing of these states by using integrated quantum frequency combs and fiber-based devices, respectively. We finally report an experimentally practical entanglement identification and characterization technique based on witness operators that is valid for any complex photon state and provides a good compromise between experimental feasibility and noise robustness. The results reported here can pave the way toward boosting the implementation of quantum technologies in integrated and widely accessible photonic platforms.

A Hopf physical reservoir computer

Physical reservoir computing utilizes a physical system as a computational resource. This nontraditional computing technique can be computationally powerful, without the need of costly training. Here, a Hopf oscillator is implemented as a reservoir computer by using a node-based architecture; however, this implementation does not use delayed feedback lines. This reservoir computer is still powerful, but it is considerably simpler and cheaper to implement as a physical Hopf oscillator. A non-periodic stochastic masking procedure is applied for this reservoir computer following the time multiplexing method. Due to the presence of noise, the Euler–Maruyama method is used to simulate the resulting stochastic differential equations that represent this reservoir computer. An analog electrical circuit is built to implement this Hopf oscillator reservoir computer experimentally. The information processing capability was tested numerically and experimentally by performing logical tasks, emulation tasks, and time series prediction tasks. This reservoir computer has several attractive features, including a simple design that is easy to implement, noise robustness, and a high computational ability for many different benchmark tasks. Since limit cycle oscillators model many physical systems, this architecture could be relatively easily applied in many contexts.