China 
Africa 
Explore chapters and articles related to this topic
From rigid automation to flexible autonomy
Published in Guy André Boy, Human–Systems Integration, 2020
In any case, automation (software) should be reliable at any time in order to support safe, efficient, and comfortable work. There are many ways to test software reliability (Lyu, 1995; Rook, 1990). In this book, what we try to promote is not only system reliability but also HSI reliability. We know that there is a co-adaptation of people and machines (via designers and engineers, as well as trainers and accumulated experience). Human operators may accept some unreliable situations where the machine may fail as long as safety, efficiency, and comfort costs are not too high (i.e., acceptable degraded modes of operations). However, when these costs become too high for them, the machine is just rejected. Again, this states the problem of product maturity (Boy, 2005); the conventional capacity maturity model for software development (Paulk et al., 1995), systematically used in most industries, does not guarantee product maturity but manufacturing process maturity. Product maturity requires continuous investment of end users in design and development processes. At the very beginning, they must be involved with domain specialists to set up high-level requirements right; this is an important role of participatory design. During the design and development phase, formative evaluations should be performed involving appropriate potential end users in order to incrementally “invent” and discover the most appropriate future use of the product in an agile way.
Biological Responses in Context
Published in Arthur T. Johnson, Biology for Engineers, 2019
There are many other examples of adaptation, from the kind and amount of food ingested in relation to food availability, to BU movement to avoid danger. In other sections, we will consider the evolution and co-adaptation of several BU in close proximity. In all cases, it should be clear that BU are not passive puppets that must be satisfied with whatever they receive from their surroundings; they are all active players that change themselves to better exploit that which is available.
A Human-Centered Design Approach
Published in Guy A. Boy, The Handbook of Human-Machine Interaction, 2017
In any case, software should be reliable at any time in order to support safe, efficient and comfortable work. There are many ways to test software reliability (Lyu, 1995; Rook, 1990). In this handbook, what we try to promote is not only system reliability, but also human–machine reliability. We know that there is a co-adaptation of people and machines (via designers and engineers). Human operators may accept some unreliable situations where the machine may fail as long as safety, efficiency and comfort costs are not too high. However, when these costs become high enough for them, the machine is just rejected. Again this poses the problem of product maturity (Boy, 2005); the conventional capacity maturity model for software development (Paulk et al., 1995), systematically used in most industries, does not guarantee product maturity, but process maturity. Product maturity requires continuous investment of end-users in design and development processes. At the very beginning, they must be involved with domain specialists to set up high-level requirements right; this is an important role of participatory design. During the design and development phase, formative evaluations should be performed involving appropriate potential end-users in order to “invent” the most appropriate future use of the product.
A Bayesian deep learning framework for interval estimation of remaining useful life in complex systems by incorporating general degradation characteristics
Published in IISE Transactions, 2020
In order to overcome these drawbacks, Gal and Ghahramani (2016) recently developed Monte Carlo dropout (MC dropout), which combines approximate Bayesian NN inference with dropout. Dropout is a stochastic regularization technique widely used to prevent co-adaptation and overfitting in NNs (Hinton et al., 2012). The key idea of dropout is to randomly remove (drop) some neurons and their connections from the NN during each iteration of stochastic gradient descent. At the test time, all neurons are present, and the trained weights are multiplied by the probability of keeping (not dropped) neurons. In contrast with conventional dropouts, which drop neurons only at training time, MC dropout trains a model with dropout and also performs a dropout at the test time. In this way, dropout can be interpreted as a variational inference approximation under certain conditions (Gal and Ghahramani, 2016). MC dropout has drawn considerable attention recently, due to its simplicity, scalability, and computational efficiency compared with the conventional Bayesian NN approaches. In this study, we will employ MC dropout to model the uncertainty stemming from unknown model parameters.
Memory-augmented meta-learning on meta-path for fast adaptation cold-start recommendation
Published in Connection Science, 2022
Tianyuan Li, Xin Su, Wei Liu, Wei Liang, Meng-Yen Hsieh, Zhuhui Chen, XuChong Liu, Hong Zhang
The objective of a co-adaptive meta-learner is to learn global prior so that it can swiftly adjust to a new task with only sparse interactions. The co-adaptation meta-learner includes semantic-based adaptation and task-based adaptation procedures, which will be described as follows.