scholarly journals A Minimum Perturbation Theory of Deep Perceptual Learning

2021 ◽  
Author(s):  
Haozhe Shan ◽  
Haim Sompolinsky
Keyword(s):  
Perceptual Learning ◽  
Network Architecture ◽  
Brain Plasticity ◽  
Deep Structure ◽  
Causal Relation ◽  
Neuronal Response ◽  
Single Layer ◽  
Behavioral Changes ◽  
Large Space ◽  
Deep Network

AbstractPerceptual learning (PL) involves long-lasting improvement in perceptual tasks following extensive training. Such improvement has been found to correlate with modifications in neuronal response properties in early as well as late sensory cortical areas. A major challenge is to dissect the causal relation between modification of the neural circuits and the behavioral changes. Previous theoretical and computational studies of PL have largely focused on single-layer model networks, and thus did not address salient characteristics of PL arising from the multiple-staged “deep” structure of the perceptual system. Here we develop a theory of PL in a deep neuronal network architecture, addressing the questions of how changes induced by PL are distributed across the multiple stages of cortex, and how do the respective changes determine the performance in fine discrimination tasks. We prove that in such tasks, modifications of synaptic weights of early sensory areas are both sufficient and necessary for PL. In addition, optimal synaptic weights in the deep network are not unique but span a large space of solutions. We postulate that, in the brain, plasticity throughout the deep network is distributed such that the resultant perturbation on prior circuit structures is minimized. In contrast to most previous models of PL, the minimum perturbation (MP) learning does not change the network readout weights. Our results provide mechanistic and normative explanations for several important physiological features of PL and reconcile apparently contradictory psychophysical findings.

scholarly journals An aggregate method for thorax diseases classification

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Bayu Adhi Nugroho
Keyword(s):  
Network Architecture ◽  
Classification Problem ◽  
Calculation Algorithm ◽  
Training Pattern ◽  
Deep Network ◽  
Network Training ◽  
Chest X Ray ◽  
Medical Image Classification ◽  
Positive Pattern

AbstractA common problem found in real-word medical image classification is the inherent imbalance of the positive and negative patterns in the dataset where positive patterns are usually rare. Moreover, in the classification of multiple classes with neural network, a training pattern is treated as a positive pattern in one output node and negative in all the remaining output nodes. In this paper, the weights of a training pattern in the loss function are designed based not only on the number of the training patterns in the class but also on the different nodes where one of them treats this training pattern as positive and the others treat it as negative. We propose a combined approach of weights calculation algorithm for deep network training and the training optimization from the state-of-the-art deep network architecture for thorax diseases classification problem. Experimental results on the Chest X-Ray image dataset demonstrate that this new weighting scheme improves classification performances, also the training optimization from the EfficientNet improves the performance furthermore. We compare the aggregate method with several performances from the previous study of thorax diseases classifications to provide the fair comparisons against the proposed method.

Reliability Models of GSM-R Redundant Network on High-Speed Railway

2012 ◽  
Vol 198-199 ◽  
pp. 1783-1788
Author(s):  
Jun Ting Lin ◽  
Jian Wu Dang
Keyword(s):  
Network Architecture ◽  
High Speed ◽  
Failure Time ◽  
Network Reliability ◽  
Single Layer ◽  
Double Layers ◽  
High Speed Railway ◽  
Reliability Models ◽  
Single Layer Network ◽  
Wireless Architectures

As a dedicated digital mobile communication system designed for railway application, GSM-R must provide reliable bidirectional channel for transmitting security data between trackside equipments and on-train computer on high-speed railways. To ensure the safety of running trains, redundant network architecture is commonly used to guarantee the reliability of GSM-R. Because of the rigid demands of railway security, it is important to build reliability mathematical models, predict the network reliability and select a suitable one. Two common GSM-R wireless architectures, co-sited double layers network and intercross single layer network, are modeled and contrasted in this paper. By calculating the reliabilities of each reliable model, it is clear that more redundant the architecture is, more reliable the system will be, the whole system will bear a less failure time per year as the benefit. Meanwhile, as the redundancy of GSM-R system raises, its equipment and maintenance will cost much, but the reliability raise gently. From the standpoint of transmission system interruption and network equipment failure, the reliability of co-sited double layer network architecture is higher than the intercross single layer one, while the viability and cost of the intercross redundant network is better than co-sited one in natural disasters such as flood and lightning. Taking fully into account reliability, viability and cost, we suggest that intercross redundant network should be chosen on high-speed railway.

scholarly journals T–S Fuzzy Model Based Multi-Branch Deep Network Architecture

IEEE Access ◽  
2020 ◽  
Vol 8 ◽  
pp. 155039-155046
Author(s):  
Faguang Wang ◽  
Yue Wang ◽  
Hongmei Wang ◽  
Chaogang Tang
Keyword(s):  
Network Architecture ◽  
Fuzzy Model ◽  
Deep Network ◽  
Model Based

scholarly journals Understanding unconventional preprocessors in deep convolutional neural networks for face identification

2019 ◽  
Vol 1 (11) ◽  
Author(s):  
Chollette C. Olisah ◽  
Lyndon Smith
Keyword(s):  
Neural Networks ◽  
Convolutional Neural Networks ◽  
Network Architecture ◽  
Data Augmentation ◽  
Color Space ◽  
Back Propagation ◽  
Face Identification ◽  
Deep Convolutional Neural Networks ◽  
Softmax Classifier ◽  
Deep Network

Abstract Deep convolutional neural networks have achieved huge successes in application domains like object and face recognition. The performance gain is attributed to different facets of the network architecture such as: depth of the convolutional layers, activation function, pooling, batch normalization, forward and back propagation and many more. However, very little emphasis is made on the preprocessor’s module of the network. Therefore, in this paper, the network’s preprocessing module is varied across different preprocessing approaches while keeping constant other facets of the deep network architecture, to investigate the contribution preprocessing makes to the network. Commonly used preprocessors are the data augmentation and normalization and are termed conventional preprocessors. Others are termed the unconventional preprocessors, they are: color space converters; grey-level resolution preprocessors; full-based and plane-based image quantization, Gaussian blur, illumination normalization and insensitive feature preprocessors. To achieve fixed network parameters, CNNs with transfer learning is employed. The aim is to transfer knowledge from the high-level feature vectors of the Inception-V3 network to offline preprocessed LFW target data; and features is trained using the SoftMax classifier for face identification. The experiments show that the discriminative capability of the deep networks can be improved by preprocessing RGB data with some of the unconventional preprocessors before feeding it to the CNNs. However, for best performance, the right setup of preprocessed data with augmentation and/or normalization is required. Summarily, preprocessing data before it is fed to the deep network is found to increase the homogeneity of neighborhood pixels even at reduced bit depth which serves for better storage efficiency.

A novel deep network architecture for reconstructing RGB facial images from thermal for face recognition

2019 ◽  
Vol 78 (18) ◽  
pp. 25259-25271 ◽  
Author(s):  
Andre Litvin ◽  
Kamal Nasrollahi ◽  
Sergio Escalera ◽  
Cagri Ozcinar ◽  
Thomas B. Moeslund ◽  
...  
Keyword(s):  
Face Recognition ◽  
Network Architecture ◽  
Deep Network ◽  
Facial Images

PanNet: A Deep Network Architecture for Pan-Sharpening

2017 ◽  
Author(s):  
Junfeng Yang ◽  
Xueyang Fu ◽  
Yuwen Hu ◽  
Yue Huang ◽  
Xinghao Ding ◽  
...  
Keyword(s):  
Network Architecture ◽  
Deep Network

scholarly journals Convolutional models of RNA energetics

2018 ◽  
Author(s):  
Michelle J. Wu
Keyword(s):  
Nucleic Acid ◽  
Network Architecture ◽  
Nearest Neighbor ◽  
Current Model ◽  
Mrna Translation ◽  
Large Space ◽  
Dna And Rna ◽  
Nucleic Acid Interactions

AbstractNucleic acid molecular biology and synthetic biology are undergoing rapid advances with the emergence of designer riboswitches controlling living cells, CRISPR/Cas9-based genome editing, high-throughput RNA-based silencing, and reengineering of mRNA translation. Many of these efforts require the design of nucleic acid interactions, which relies on accurate models for DNA and RNA energetics. Existing models utilize nearest neighbor rules, which were parameterized through careful optical melting measurements. However, these relatively simple rules often fail to quantitatively account for the biophysical behavior of molecules even in vitro, let alone in vivo. This is due to the limited experimental throughput of optical melting experiments and the infinitely large space of possible motifs that can be formed. Here, we present a convolutional neural network architecture to model the energies of nucleic acid motifs, allowing for learning of representations of physical interactions that generalize to arbitrary unmeasured motifs. First, we used existing parameterizations of motif energies to train the model and demonstrate that our model is expressive enough to recapitulate the current model. Then, through training on optical melting datasets from the literature, we have shown that the model can accurately predict the thermodynamics of hairpins containing unmeasured motifs. This work demonstrates the utility of convolutional models for capturing the thermodynamic parameters that underlie nucleic acid interactions.

scholarly journals Exogenous attention generalizes perceptual learning in adults with amblyopia

2021 ◽  
Author(s):  
Mariel Roberts ◽  
Marisa Carrasco
Keyword(s):  
Perceptual Learning ◽  
Brain Plasticity ◽  
Visual Development ◽  
Adult Brain ◽  
Exogenous Attention ◽  
Visual Perceptual Learning ◽  
Behavioral Manifestation ◽  
Improved Performance ◽  
Spatial Locations ◽  
Insight Into

SUMMARYVisual perceptual learning (VPL), or improved performance after practicing the same visual task, is a behavioral manifestation of the impressive neuroplasticity in the adult brain. However, its practical effectiveness is limited because improvements are often specific to the trained conditions and require significant time and effort. Thus, it is critical to understand the conditions that promote learning and its transfer. Covert spatial attention helps overcome VPL location and feature specificity in neurotypical adults, but whether it can for people with atypical visual development is unknown. Here we show that involuntary attention helps generalize learning beyond trained spatial locations in adults with amblyopia, an ideal population for investigation given their asymmetrically developed, but highly plastic, visual cortex. Our findings provide insight into the mechanisms underlying changes in neuro(a)typical brain plasticity after practice. Further, they reveal that attention can enhance the effectiveness of perceptual learning during rehabilitation of visual disorders.

Sign in / Sign up

Export Citation Format

Share Document