2.2.4. Visualizing the resultant functions of BO When d ≤ 2, the resultant functions of BO are displayed as two-dimensional (2D) graphs. If d=2, the user can change the graph-type from 2D heatmaps to three-dim
The models in [3,4] are also similar to our first stage optimisation in terms of the security risk, and the direct and indirect costs in the objective functions, but they only consider single-step attacks instead of multi-step attack scenarios. Other well known “single-step” approaches ...
Unfortunately, most BNSL algorithms use local optimisation methods, which do not guarantee the optimal Bayesian network structure in a global sense. In recent years, a method has been developed to discover the optimal structure of Bayesian networks globally. This method based on Integer Linear ...
The choice of model boundaries, the level of abstraction or idealisation, mathematical representation and used optimisation routines are highly individual for every energy model. Early attempts of quality improvement for energy models have mainly focussed on technical modelling aspects [6]. Recent ...
A method for nondifferentiable optimization in MAP (maximum a posteriori) estimation of computed tomographs is presented. This problem arises in the application of a Markov random field image model with absolute value potential function. The algorithm uses local optimization operations, operating in alt...
This Section gives some details on the optimisation phase: both the optimisation of the availability and of the robustness are characterised by the same “meta-schema”. For the sake of clarity, we illustrate these techniques on the BN model in Fig. 10. According to the workflow in Fig. 3...
The choice of model boundaries, the level of abstraction or idealisation, mathematical representation and used optimisation routines are highly individual for every energy model. Early attempts of quality improvement for energy models have mainly focussed on technical modelling aspects [6]. Recent ...
,N} is the number of poles. Please note that the parameter set θ can only obtain non-negative values. Searching for ϕ using (5) is affected by two main problems that occur during the optimisation process: the degeneration of solutions and the range of parameters’ values. The ...
Structure learning methods generally fall into two main classes of learning known as score-based and constraint-based learning. The score-based algorithms rely on search methods that explore the search space of graphs and an objective function that scores each graph visited, where the highest ...
The selection of the hyper-parameters as well as the neural network architecture is done via a grid search by varying all parameters such as the optimisation algorithm and the learning rate listed in Table 1. We deploy the ML surrogate model MML to compute 15,000 input–output pairs (linking...