Transmission electron microscopy of high temperature annealing of pure tungsten irradiated by self-ions was conducted to elucidate microstructural and defect evolution in temperature ranges relevant to fusion reactor applications (500-1200°C). Bulk isochronal and isothermal annealing of ion irradiated pure tungsten (2 MeV W + ions, 500°C, 10 14 W + /cm 2 ) with temperatures of 800, 950, 1100 and 1400°C, from 0.5 to 8 h, was followed by ex situ characterisation of defect size, number density, Burgers vector and nature. Loops with diameters larger than 2-3 nm were considered for detailed analysis, among which all loops had b ¼ 1 2 h1 1 1i and were predominantly of interstitial nature. In situ annealing experiments from 300 up to 1200°C were also carried out, including dynamic temperature ramp-ups. These confirmed an acceleration of loop loss above 900°C. At different temperatures within this range, dislocations exhibited behaviour such as initial isolated loop hopping followed by large-scale rearrangements into loop chains, coalescence and finally line-loop interactions and widespread absorption by free-surfaces at increasing temperatures. An activation energy for the annealing of dislocation length was derived, finding E a ¼ 1:34 AE 0:2 eV for the 700-1100°C range.

We present a sampling-free approach for computing the epistemic uncertainty of a neural network. Epistemic uncertainty is an important quantity for the deployment of deep neural networks in safety-critical applications, since it represents how much one can trust predictions on new data. Recently promising works were proposed using noise injection combined with Monte-Carlo (MC) sampling at inference time to estimate this quantity (e.g. MC dropout). Our main contribution is an approximation of the epistemic uncertainty estimated by these methods that does not require sampling, thus notably reducing the computational overhead. We apply our approach to large-scale visual tasks (i.e., semantic segmentation and depth regression) to demonstrate the advantages of our method compared to sampling-based approaches in terms of quality of the uncertainty estimates as well as of computational overhead.

The length of the geodesic between two data points along a Riemannian manifold, induced by a deep generative model, yields a principled measure of similarity. Current approaches are limited to low-dimensional latent spaces, due to the computational complexity of solving a non-convex optimisation problem. We propose finding shortest paths in a finite graph of samples from the aggregate approximate posterior, that can be solved exactly, at greatly reduced runtime, and without a notable loss in quality. Our approach, therefore, is hence applicable to high-dimensional problems, e.g., in the visual domain. We validate our approach empirically on a series of experiments using variational autoencoders applied to image data, including the Chair, FashionMNIST, and human movement data sets.

The ability to steer flexible needles and probes to access deep anatomical locations safely for medical diagnosis and treatment represents a current clinical and engineering research challenge. The behaviour of parasitic wasps has inspired the development of a novel steerable and flexible multi-part probe, which allows the control of its approach angle by adjusting the steering offset between probe segments, i.e. by means of a programmable bevel tip. This paper describes the experimental evaluation of several scaled-up proof-of-concept flexible probe prototypes to explore the effects of tip design (bevel-tip angle) and dimensions (outer diameter) on steering. For each prototype, a linear relationship between steering offset and curvature is confirmed. The effect of probe diameter and bevel-tip angle on steering performance is also analysed, with results confirming that larger bevel-tip angles and smaller probe diameters lead to larger curvature values, although improved steering comes at the price of a less stable insertion process.

No abstract