Larsen, Thomas NakkenMeyer, EivindRasheed, AdilSan, Omer...
17页
查看更多>>摘要：Autonomous systems are becoming ubiquitous and gaining momentum within the marine sector. Since the electrification of transport is happening simultaneously, autonomous marine vessels can reduce environmental impact, lower costs, and increase efficiency. Although close monitoring is still required to ensure safety, the ultimate goal is full autonomy. One major milestone is to develop a control system that is versatile enough to handle any weather and encounter that is also robust and reliable. Additionally, the control system must adhere to the International Regulations for Preventing Collisions at Sea (COLREGs) for successful interaction with human sailors. Since the COLREGs were written for the human mind to interpret, they are written in ambiguous prose and therefore not machine-readable or verifiable. Due to these challenges and the wide variety of situations to be tackled, classical model-based approaches prove complicated to implement and computationally heavy. Within machine learning (ML), deep reinforcement learning (DRL) has shown great potential for a wide range of applications. The model-free and self-learning properties of DRL make it a promising candidate for autonomous vessels. In this work, a subset of the COLREGs is incorporated into a DRL-based path following and obstacle avoidance system using collision risk theory. The resulting autonomous agent dynamically interpolates between path following and COLREG-compliant collision avoidance in the training scenario, isolated encounter situations, and AIS-based simulations of real-world scenarios. (C) 2022 The Author(s). Published by Elsevier Ltd.
Solis-Perez, J. E.Hernandez, J. A.Parrales, A.Gomez-Aguilar, J. F....
13页
查看更多>>摘要：This research proposes a novel transfer function based on the hyperbolic tangent and the Khalil conformable exponential function. The non-integer order transfer function offers a suitable neural network configuration because of its ability to adapt. Consequently, this function was introduced into neural network models for three experimental cases: estimating the annular Nusselt number correlation to a helical double-pipe evaporator, the volumetric mass transfer coefficient in an electrochemical reaction, and the thermal efficiency of a solar parabolic trough collector. We found the new transfer function parameters during the training step of the neural networks. Therefore, weights and biases depend on them. We assessed the models applied to the three cases using the determination coefficient, adjusted determination coefficient, and the slope-intercept test. In addition, the MSE for the training set and the whole database were computed to show that there is no overfitting problem. The best-assessed models showed a relationship of 99%, 97%, and 95% with the experimental data for the first, second, and third cases. This novel proposal made reducing the number of neurons in the hidden layer feasible. Therefore, we show a neural network with a conformable transfer function (ANN-CTF) that learns well enough with less available information from the experimental database during its training.
查看更多>>摘要：Two-photon fluorescence microscopy has enabled the three-dimensional (3D) neural imaging of deep cortical regions. While it can capture the detailed neural structures in the x-y image space, the image quality along the depth direction is lower because of lens blur, which often makes it difficult to identify the neural connectivity. To address this problem, we propose a novel approach for restoring the isotropic image volume by estimating and fusing the intersection regions of the images captured from three orthogonal viewpoints using convolutional neural networks (CNNs). Because convolution on 3D images is computationally complex, the proposed method takes the form of cascaded CNN models consisting of rigid transformation, dense registration, and deblurring networks for more efficient processing. In addition, to enable self-supervised learning, we trained the CNN models with simulated synthetic images by considering the distortions of the microscopic imaging process. Through extensive experiments, the proposed method achieved substantial image quality improvements.(c) 2022 Elsevier Ltd. All rights reserved.
Ben-Iwhiwhu, EseogheneDick, JefferyKetz, Nicholas A.Pilly, Praveen K....
10页
查看更多>>摘要：Meta-reinforcement learning (meta-RL) algorithms enable agents to adapt quickly to tasks from few samples in dynamic environments. Such a feat is achieved through dynamic representations in an agent's policy network (obtained via reasoning about task context, model parameter updates, or both). However, obtaining rich dynamic representations for fast adaptation beyond simple benchmark problems is challenging due to the burden placed on the policy network to accommodate different policies. This paper addresses the challenge by introducing neuromodulation as a modular component to augment a standard policy network that regulates neuronal activities in order to produce efficient dynamic representations for task adaptation. The proposed extension to the policy network is evaluated across multiple discrete and continuous control environments of increasing complexity. To prove the generality and benefits of the extension in meta-RL, the neuromodulated network was applied to two state-of-the-art meta-RL algorithms (CAVIA and PEARL). The result demonstrates that meta-RL augmented with neuromodulation produces significantly better result and richer dynamic representations in comparison to the baselines. (C) 2022 The Authors. Published by Elsevier Ltd.
查看更多>>摘要：This paper studies the multistability of delayed recurrent neural networks (DRNNs) with a class of piecewise nonlinear activation functions. The coexistence as well as the stability of multiple equilibrium points (EPs) of DRNNs are proved. With the Brouwer's fixed point theorem as well as the Lagrange mean value theorem, it is obtained that under some conditions, the n-neuron DRNNs with the proposed activation function can have at least 5(n) EPs and 3(n) of them are locally stable. Compared with the DRNNs with sigmoidal activation functions, DRNNs with this kind of activation function can have more total EPs and more locally stable EPs. It implies that when designing DRNNs with the proposed activation function to apply in associative memory, it can have an even larger storage capacity. Furthermore, it is obtained that there exists a relationship between the number of the total EPs/stable EPs and the frequency of the sinusoidal function in the proposed activation function. Last, the above obtained results are extended to a more general case. It is shown that, DRNNs with the extended activation function can have (2k + 1)(n) EPs, (k + 1)(n) of which are locally stable, therein k is closely related to the frequency of the sinusoidal function in the extended activation function. Two simulation examples are given to verify the correctness of the theoretical results. (C) 2022 Elsevier Ltd. All rights reserved.
查看更多>>摘要：Reinforcement learning algorithms are typically limited to learning a single solution for a specified task, even though diverse solutions often exist. Recent studies showed that learning a set of diverse solutions is beneficial because diversity enables robust few-shot adaptation. Although existing methods learn diverse solutions by using the mutual information as unsupervised rewards, such an approach often suffers from the bias of the gradient estimator induced by value function approximation. In this study, we propose a novel method that can learn diverse solutions without suffering the bias problem. In our method, a policy conditioned on a continuous or discrete latent variable is trained by directly maximizing the variational lower bound of the mutual information, instead of using the mutual information as unsupervised rewards as in previous studies. Through extensive experiments on robot locomotion tasks, we demonstrate that the proposed method successfully learns an infinite set of diverse solutions by learning continuous latent variables, which is more challenging than learning a finite number of solutions. Subsequently, we show that our method enables more effective few-shot adaptation compared with existing methods. (C) 2022 Elsevier Ltd. All rights reserved.
查看更多>>摘要：Artificial neural network has been fully developed in recent years, but as the size of the network grows, the required computing power also grows rapidly. In order to take advantage of the parallel computing of quantum computing to solve the difficulties of large computation in neural network, quantum neural network was proposed. In this paper, based on the pulse coupled neural network (PCNN), quantum pulse coupled neural network (QPCNN) is proposed. In this model, the basic quantum logic gates are utilized to form quantum operation modules, such as quantum full adder, quantum multiplier, and quantum comparator. A quantum image convolution operation applicable to QPCNN is designed employing quantum full adders and neighborhood preparation module. And these modules are employed to complete the operations required for QPCNN. And based on QPCNN, an quantum image segmentation is designed. Meanwhile, the effectiveness of QPCNN is proved by simulation experiments, and the complexity analysis shows that QPCNN has exponential speedup compared with classical PCNN. (C) 2022 Elsevier Ltd. All rights reserved.
查看更多>>摘要：This paper aims at solving a stochastic two-player zero-sum Nash game problem studied in Singh and Lisser (2019). The main contribution of our paper is that we model this game problem as a dynamical neural network (DNN for short). In this paper, we show that the saddle point of this game problem is the equilibrium point of the DNN model, and we study the globally asymptotically stable of the DNN model. In our numerical experiments, we present the time-continuous feature of the DNN model and compare it with the state-of-the-art convex solvers, i.e., Splitting conic solver (SCS for short) and Cvxopt. Our numerical results show that our DNN method has two advantages in dealing with this game problem. Firstly, the DNN model can converge to a better optimal point. Secondly, the DNN method can solve all problems, even when the problem size is large. (C) 2022 Elsevier Ltd. All rights reserved.
Bekhouche, S. E.Kajo, I.Ruichek, Y.Dornaika, F....
10页
查看更多>>摘要：Eye blink detection is a challenging problem that many researchers are working on because it has the potential to solve many facial analysis tasks, such as face anti-spoofing, driver drowsiness detection, and some health disorders. There have been few attempts to detect blinking in the wild scenario, while most of the work has been done under controlled conditions. Moreover, current learning approaches are designed to process sequences that contain only a single blink ignoring the case of the presence of multiple eye blinks. In this work, we propose a fast framework for eye blink detection and eye blink verification that can effectively extract multiple blinks from image sequences considering several challenges such as lighting changes, variety of poses, and change in appearance. The proposed framework employs fast landmarks detector to extract multiple facial key points including the ones that identify the eye regions. Then, an SVD-based method is proposed to extract the potential eye blinks in a moving time window that is updated with new images every second. Finally, the detected blink candidates are verified using a 2D Pyramidal Bottleneck Block Network (PBBN). We also propose an alternative approach that uses a sequence of frames instead of an image as input and employs a continuous 3D PBBN that follows most of the state-of-the-art approaches schemes. Experimental results show the better performance of the proposed approach compared to the state-of-the-art approaches.(c) 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
查看更多>>摘要：Square matrices appear in many machine learning problems and models. Optimization over a large square matrix is expensive in memory and in time. Therefore an economic approximation is needed. Conventional approximation approaches factorize the square matrix into a number matrices of much lower ranks. However, the low-rank constraint is a performance bottleneck if the approximated matrix is intrinsically high-rank or close to full rank. In this paper, we propose to approximate a large square matrix with a product of sparse full-rank matrices. In the approximation, our method needs only N(log N)(2) non-zero numbers for an N x N full matrix. Our new method is especially useful for scalable neural attention modeling. Different from the conventional scaled dot-product attention methods, we train neural networks to map input data to the non-zero entries of the factorizing matrices. The sparse factorization method is tested for various square matrices, and the experimental results demonstrate that our method gives a better approximation when the approximated matrix is sparse and high rank. As an attention module, our new method defeats Transformer and its several variants for long sequences in synthetic data sets and in the Long Range Arena benchmarks. Our code is publicly available(2). (C) 2022 The Author(s) .Published by Elsevier Ltd.
NSTL
Elsevier