Most of the current knowledge management approaches deal with the problems of knowledge identification and formalization. In some case, and especially in the knowledge engineering field, they deal with the problems of...
详细信息
Most of the current knowledge management approaches deal with the problems of knowledge identification and formalization. In some case, and especially in the knowledge engineering field, they deal with the problems of knowledge reuse modeling. However there are few researches dealing with both themes of knowledge management and skills management to make the reuse of knowledge easier and more valuable to people. We present in this paper an organizational Meta-Model which aims to highlight the links between these two research themes. Based on our results with an Ophthalmic industry, we will present you our approach to define KROM (Knowledge Reuse Organizational Meta-Model) an organizational model supporting the knowledge reuse.
As an important issue in signal processing field, filter design is essentially a multiple-parameter optimization problem. Because the searching process of pure simulated annealing is rather long, and pure genetic is e...
详细信息
As an important issue in signal processing field, filter design is essentially a multiple-parameter optimization problem. Because the searching process of pure simulated annealing is rather long, and pure genetic is easy to be premature convergent, combining the probabilistic jumping search ability of simulated annealing with genetic fast converge to some local minimum of the search space, this paper proposes an effective and easy-to-be implemented parallel annealing-genetic strategy for soft morphological filters design. According to the empirical results as well as comparison with conventional genetic and simulated annealing algorithms, the effective and global optimization ability of the proposed strategy are verified.
A four terminal Gaussian network composed of a source, a destination, an eavesdropper and a jammer relay is investigated when the jammer relay is causally given the source message. The source aims to increase the achi...
详细信息
ISBN:
(纸本)9781424480166
A four terminal Gaussian network composed of a source, a destination, an eavesdropper and a jammer relay is investigated when the jammer relay is causally given the source message. The source aims to increase the achievable secrecy rates, whereas the jammer relay aims to decrease it. To help the eavesdropper and to decrease achievable perfect secrecy rates, the jammer relay can use pure relaying and/or send interference to assist eavesdropping. The problem is formulated as a zero-sum game and the saddle point solutions are found. The results are compared to the case when the jammer relay is not informed about the source message.
This paper proposes a method to regulate the rotational speed of an induction motor (IM) using model predictive direct torque control (MPDTC) and Fuzzy Gain Scheduling of PI controller (FGS-PI). MPDTC uses a cost func...
详细信息
To make reinforcement learning more sample efficient, we need better credit assignment methods that measure an action's influence on future rewards. Building upon Hindsight Credit Assignment (HCA) [1], we introduc...
详细信息
ISBN:
(纸本)9781713899921
To make reinforcement learning more sample efficient, we need better credit assignment methods that measure an action's influence on future rewards. Building upon Hindsight Credit Assignment (HCA) [1], we introduce Counterfactual Contribution Analysis (COCOA), a new family of model-based credit assignment algorithms. Our algorithms achieve precise credit assignment by measuring the contribution of actions upon obtaining subsequent rewards, by quantifying a counterfactual query: 'Would the agent still have reached this reward if it had taken another action?'. We show that measuring contributions w.r.t. rewarding states, as is done in HCA, results in spurious estimates of contributions, causing HCA to degrade towards the high-variance REINFORCE estimator in many relevant environments. Instead, we measure contributions w.r.t. rewards or learned representations of the rewarding objects, resulting in gradient estimates with lower variance. We run experiments on a suite of problems specifically designed to evaluate long-term credit assignment capabilities. By using dynamic programming, we measure ground-truth policy gradients and show that the improved performance of our new model-based credit assignment methods is due to lower bias and variance compared to HCA and common baselines. Our results demonstrate how modeling action contributions towards rewarding outcomes can be leveraged for credit assignment, opening a new path towards sample-efficient reinforcement learning.(2)
暂无评论