No abstract

Magnetostrictive materials can be used to construct high bandwidth actuators with a higher force and a larger stroke than are provided by other materials. However, their use is hindered by their complex nonlinear and hysteretic response. This response displays a significant dependence on mechanical loading. In this paper, a modeling technique is introduced for reproducing hysteresis curves at different loads. The classic Preisach model is used, although the approach can be used to include load dependence in other models. Predicted values are compared with the homogenized energy model and also with experimental data.

How students experience educational environments and the interconnections between their readiness, task experiences and their long-term desire to reengage with course content are critical questions for educators. Research postgraduate students (n=310) at a research- intensive university in Hong Kong, engaging in a 24-hour introductory teaching course, participated in this study. Learner readiness for the course was assessed as prior Domain interest, self-efficacy, and knowledge. Subsequently, students completed four formative assessments, reported their on-task interest in seven strategically chosen tasks and end-of- course Course and Domain interest. Longitudinal-SEM tested interconnections between readiness components, Task, Course and Domain interest. Initial self-efficacy beliefs for teaching predicted early Task interest, while Domain interest was a predictor of Task interest in explicitly practical task experiences. Strong interconnections between Task interest across the study were evident. Individual written and social (discussion) tasks presented strong contributions to future Course/Domain interest. Implications for theory and practice are discussed.

The Ordered Upwind Method (OUM) is used to approximate the viscosity solution of the static Hamilton-Jacobi-Bellman (HJB) with direction-dependent weights on unstructured meshes. The method has been previously shown to provide a solution that converges to the exact solution, but no convergence rate has been theoretically proven. In this paper, it is shown that the solutions produced by the OUM in the boundary value formulation converge at a rate of at least the square root of the largest edge length in the mesh in terms of maximum error. An example with similar order of numerical convergence is provided

Pretrained language models have served as the backbone for many state-of-the-art NLP results. These models are large and expensive to train. Recent work suggests that continued pretraining on task-specific data is worth the effort as pretraining leads to improved performance on downstream tasks. We explore alternatives to full-scale task-specific pretraining of language models through the use of adapter modules, a parameter-efficient approach to transfer learning. We find that adapter-based pretraining is able to achieve comparable results to task-specific pretraining while using a fraction of the overall trainable parameters. We further explore direct use of adapters without pretraining and find that the direct finetuning performs mostly on par with pretrained adapter models, contradicting previously proposed benefits of continual pretraining in full pretraining fine-tuning strategies. Lastly, we perform an ablation study on task-adaptive pretraining to investigate how different hyperparameter settings can change the effectiveness of the pretraining.