课程文献中心 更多>>

摘要： The plant cell wall is an extracellular matrix that envelopes cells, gives them structure and shape, constitutes the interface with symbionts, and defends plants against external biotic and abiotic stress factors. The assembly of this matrix is regulated and mediated by the cytoskeleton. Cytoskeletal elements define where new cell wall material is added and how fibrillar macromolecules are oriented in the wall. Inversely, the cytoskeleton is also key in the perception of mechanical cues generated by structural changes in the cell wall as well as the mediation of intracellular responses. We review the delivery processes of the cell wall precursors that are required for the cell wall assembly process and the structural continuity between the inside and the outside of the cell. We provide an overview of the different morphogenetic processes for which cell wall assembly is a crucial element and elaborate on relevant feedback mechanisms.

关键词： Time series classification   Learning systems   Capsule networks   Data mining   Multi-head convolutional neural networks   Signal processing  

摘要： Time series classification (TSC) is the problem of categorizing time series data by using machine learning techniques. Its applications vary from cybersecurity and health care to remote sensing and human activity recognition. In this paper, we propose a novel multi-process collaborative architecture for TSC. The propositioned method amalgamates multi-head convolutional neural networks and capsule mechanism. In addition to the discovery of the temporal relationship within time series data, our approach derives better feature extraction with different scaled capsule routings and enhances representation learning. Unlike the original CapsNet, our proposed approach does not need to reconstruct to increase the accuracy of the model. We examine our proposed method through a set of experiments running on the domain-agnostic TSC benchmark datasets from the UCR Time Series Archive. The results show that, compared to a number of recently developed and currently used algorithms, we achieve 36 best accuracies out of 128 datasets. The accuracy analysis of the proposed approach demonstrates its significance in TSC by offering very high classification confidence with the potential of making inroads into plentiful future applications. (c) 2021 Elsevier B.V. All rights reserved.

摘要： Reality is noisy and messy, and there is no grand simulator of things in sight. What is more, our models can only restrict their scope to increasingly smaller empirical nooks, solving, at best, tiny fractions of an infnite jigsaw puzzle. As we, researchers in need of cognition, go through the process of building new models and discarding old ones, we require the right tools to foster our epistemic achievements. At the time of writing, deep learning continues to enjoy a vibrant hype, refurbishing the methodological equipment of many quantitative sciences. The behavioral sci- ences, despite being more resistant to change than fellow disciplines, are also enjoying their fair share of the rapidly expanding trend. When it comes to model-based inference, deep learning innovations are currently transforming the way models are ft to data and employed for draw- ing substantive conclusions or deriving reliable forecasts. Moreover, uncertainty (an ancient con- cept) and its quantifcation are becoming more and more important in deep learning theory and practice. Bayesian methods, deeply rooted in probability theory, are currently viewed by many researchers as the gold-standard for uncertainty-aware inference, but other approaches or gener- alizations might push through in the not-so-distant future. In a way, this thesis ventured into a discourse between deep learning and scientifc modeling with a focus on cognitive science and mathematical psychology. It brought together ideas for dramatically accelerating Bayesian infer- ence by using non-Bayesian neural networks designed to deal with the data types encountered by researchers working in various areas of knowledge. The general idea of using black-box estimators to learn white-box scientifc models from computer simulations is certainly not new, but is still largely underutilized in the behavioral and cognitive sciences. Most importantly, future research should further foster the discourse between deep learning (or artificial int

关键词： Terra-cotta   Engineering standards