Some forms of novel visual media enable the viewer to explore a 3D scene from arbitrary viewpoints, by interpolating between a discrete set of original views. Compared to 2D imagery, these types of applications require much larger amounts of storage
space, which we seek to reduce. Existing approaches for compressing 3D scenes are based on a separation of compression and rendering: each of the original views is compressed using traditional 2D image formats; the receiver decompresses the views and then performs the rendering. We unify these steps by directly compressing an implicit representation of the scene, a function that maps spatial coordinates to a radiance vector field, which can then be queried to render arbitrary viewpoints. The function is implemented as a neural network and jointly trained for reconstruction as well as compressibility, in an end-to-end manner, with the use of an entropy penalty on the parameters. Our method significantly outperforms a state-of-the-art conventional approach for scene compression, achieving simultaneously higher quality reconstructions and lower bitrates. Furthermore, we show that the performance at lower bitrates can be improved by jointly representing multiple scenes using a soft form of parameter sharing.
Variational Optimization forms a differentiable upper bound on an objective. We show that approaches such as Natural Evolution Strategies and Gaussian Perturbation, are special cases of Variational Optimization in which the expectations are approxima
ted by Gaussian sampling. These approaches are of particular interest because they are parallelizable. We calculate the approximate bias and variance of the corresponding gradient estimators and demonstrate that using antithetic sampling or a baseline is crucial to mitigate their problems. We contrast these methods with an alternative parallelizable method, namely Directional Derivatives. We conclude that, for differentiable objectives, using Directional Derivatives is preferable to using Variational Optimization to perform parallel Stochastic Gradient Descent.
An overview is presented of a method to search for $D^0to{}e^{pm}mu^{mp}$ with LHCb data. In order to reduce combinatorial backgrounds, tagged $D^0$ candidates from the decay $D^{ast+}to{}D^0pi^+$ are used. This measurement is performed with respect
to $mathcal{B}left(D^0to{}pi^+pi^-right)$, which cancels uncertainties in the luminosity and $D^{ast+}$ production cross-section. It is estimated that using $3,mathrm{fb}^{-1}$ of LHCb data an upper limit can be attained of $mathcal{O}left(10^{-7}right)$ at a $90%$ confidence level.
A mechanism is presented that suggests shielded 3-D magnetic perturbations can destabilize microinstabilities and enhance the associated anomalous transport. Using local 3-D equilibrium theory, shaped tokamak equilibria with small 3-D deformations ar
e constructed. In the vicinity of rational magnetic surfaces, the infinite-n ideal MHD ballooning stability boundary is strongly perturbed by the 3-D modulations of the local magnetic shear associated with the presence of nearresonant Pfirsch-Schluter currents. These currents are driven by 3-D components of the magnetic field spectrum even when there is no resonant radial component. The infinite-n ideal ballooning stability boundary is often used as a proxy for the onset of virulent kinetic ballooning modes (KBM) and associated stiff transport. These results suggest that the achievable pressure gradient may be lowered in the vicinity of low order rational surfaces when 3-D magnetic perturbations are applied. This mechanism may provide an explanation for the observed reduction in the peak pressure gradient at the top of the edge pedestal during experiments where edge localized modes have been completely suppressed by applied 3-D magnetic fields.
