Search

Active learning is usually applied to acquire labels of informative data points in supervised learning, to maximize accuracy in a sample-efficient way. However, maximizing the accuracy is not the end goal when the results are used for decision-making, for example in personalized medicine or economics. We argue that when acquiring samples sequentially, separating learning and decision-making is sub-optimal, and we introduce an active learning strategy which takes the down-the-line decision problem into account. Specifically, we introduce a novel active learning criterion which maximizes the expected information gain on the posterior distribution of the optimal decision. We compare our targeted active learning strategy to existing alternatives on both simulated and real data, and show improved performance in decision-making accuracy.

Perturbation theory alone fails to describe thermodynamics of the electroweak phase transition. We review a technique combining perturbative and non-perturbative methods to overcome this challenge. Accordingly, the principal theme is a tutorial of high-temperature dimensional reduction. We present an explicit derivation with a real singlet scalar and compute the thermal effective potential at two-loop order. In particular, we detail the dimensional reduction for a real-singlet extended Standard Model. The resulting effective theory will impact future non-perturbative studies based on lattice simulations as well as purely perturbative investigations., Comment: 86 pages, 2 figures and 141 Feynman diagrams

In this paper, we propose two contributions to neural network based denoising. First, we propose applying separate convolutional layers to each sub-band of discrete wavelet transform (DWT) as opposed to the common usage of DWT which concatenates all sub-bands and applies a single convolution layer. We show that our approach to using DWT in neural networks improves the accuracy notably, due to keeping the sub-band order uncorrupted prior to inverse DWT. Our second contribution is a denoising loss based on top k-percent of errors in frequency domain. A neural network trained with this loss, adaptively focuses on frequencies that it fails to recover the most in each iteration. We show that this loss results into better perceptual quality by providing an image that is more balanced in terms of the errors in frequency components.

We propose a scheme for enhancing the optomechanical coupling between microwave and mechanical resonators by up to seven orders of magnitude to the ultrastrong coupling limit in a circuit optomechanical setting. The tripartite system considered here consists of a Josephson junction Cooper-pair box that mediates the coupling between the microwave cavity and the mechanical resonator. The optomechanical coupling can be modified by tuning the gate charge and the magnetic flux bias of the Cooper-pair box which in turn affect the Josephson capacitance of the Cooper-pair box. We additionally show that with suitable choice of tuning parameters, the optomechanical coupling vanishes and the system exhibits purely a cross-Kerr type of nonlinearity between the cavity and the mechanical resonator. This allows the system to be used for phonon counting.

Non-integer dimensions are commonplace in quantum field theories (QFTs) through dimensional regularization. In particular this affects angular calculations involving dot products. The structure of these rises from the generally accepted axiom that Gaussian integrals can be written as a $d$-dimensional product of a single dimensional Gaussian integral. This result can be extended in a straightforward manner to involve any above zero-dimensional "surfaces", but there is somewhat clear ambiguity with convergence when considering negative values of dimensions. This obstacle can be answered with proper regularization strategy, which leads to an acceptable analytic continuation. Furthermore, we suggest a method of symmetrizing the angular calculations back to positive dimensional variants by applying the symmetries of Euler gamma functions. Through this method, the region $d \in [0,1]$ is recognized to be fundamentally different from either remaining half-axis of dimensionality. By setting this region as the proper limit for the angle generating dimension, we fully establish the rules of iterative use of the generating method and the maximal number integration angles., Comment: 10 pages, 1 figure

Dimensional regularization of Euclidean momentum space integrals is a highly successful technique in renormalization of quantum field theories. While it yields a straightforward algorithmic method, with which to evaluate diagrams beyond tree level, the actual integrals can be highly divergent, at least in a traditional sense. In particular, standard one-loop integrals can be expressed in terms of an explicit formula, which associates both ultraviolet and infrared divergent parameter values to analytically continued special function expressions. We aim to discuss the formulation of finite integral expressions corresponding to the analytically continued structures. Effectively, we wish to establish conditions which form an equivalence class for this analytical continuation, or rather form a proper set/conditions of regularization techniques leading to it. This is further demonstrated by considering both partially and fully successful strategies side-by-side, with major emphasis on the two simplest functioning schemes: Gaussian and cut-off regularization. By explicit computations we aim to associate these generalisations of the initial integrals with the results from dimensional regularization, considering both multiple mass (or momentum) scales as well as scaleless cases. We achieve the finiteness of the sought-after integrals by applying one of these suitable schemes along with an additional scheme related scale. This enables us to devise a proper representation of the dimensionally regularized expressions (or a local description) through an operator removing all excess terms with the additional scale(s)., Comment: 19 pages; references added

Mixed reality (MR) applications are expected to become common when 5G goes mainstream. However, the latency requirements are challenging to meet due to the resources required by video-based remoting of graphics, that is, decoding video codecs. We propose an approach towards tackling this challenge: a client-server implementation for transacting intermediate representation (IR) between a mobile UE and a MEC server instead of video codecs and this way avoiding video decoding. We demonstrate the ability to address latency bottlenecks on edge computing workloads that transact graphics. We select SPIR-V compatible GPU kernels as the intermediate representation. Our approach requires know-how in GPU architecture and GPU domain-specific languages (DSLs), but compared to video-based edge graphics, it decreases UE device delay by sevenfold. Further, we find that due to low cold-start times on both UEs and MEC servers, application migration can happen in milliseconds. We imply that graphics-based location-aware applications, such as MR, can benefit from this kind of approach.

The running coupling of a generic field theory can be described through a separable differential equation involving the corresponding $\beta$-function. Only the first loop order can be solved analytically in terms of well-known functions, all further loop orders lead to transcendental equations. While obscure nowadays, many analytical methods have been devised to study them, most specifically the Lagrange-B\"urmann formula. In this article we discuss the structure of transcendental equations that take place at various loop orders. Beyond the first two loop orders, these equations are simplified by applying an optimal Pade approximant on the $\beta$-function. In general, these lead to generalizations of Lambert's equation, the solutions of which are presented in terms of a power series., Comment: 15 pages

A large body of animation research focuses on optimization of movement control, either as action sequences or policy parameters. However, as closed-form expressions of the objective functions are often not available, our understanding of the optimization problems is limited. Building on recent work on analyzing neural network training, we contribute novel visualizations of high-dimensional control optimization landscapes; this yields insights into why control optimization is hard and why common practices like early termination and spline-based action parameterizations make optimization easier. For example, our experiments show how trajectory optimization can become increasingly ill-conditioned with longer trajectories, but parameterizing control as partial target states---e.g., target angles converted to torques using a PD-controller---can act as an efficient preconditioner. Both our visualizations and quantitative empirical data also indicate that neural network policy optimization scales better than trajectory optimization for long planning horizons. Our work advances the understanding of movement optimization and our visualizations should also provide value in educational use., Comment: Accepted to IEEE Transactions on Visualization and Computer Graphics (IEEE TVCG)

We study the effects of the coupling between magnetization dynamics and the electronic degrees of freedom in a heterostructure of a metallic nanomagnet with dynamic magnetization coupled with a superconductor containing a steady spin-splitting field. We predict how this system exhibits a non-linear spin torque, which can be driven either with a temperature difference or a voltage across the interface. We generalize this notion to arbitrary magnetization precession by deriving a Keldysh action for the interface, describing the coupled charge, heat and spin transport in the presence of a precessing magnetization. We characterize the effect of superconductivity on the precession damping and the anti-damping torques. We also predict the full non-linear characteristic of the Onsager counterparts of the torque, showing up via pumped charge and heat currents. For the latter, we predict a spin-pumping cooling effect, where the magnetization dynamics can cool either the nanomagnet or the superconductor., Comment: 9 pages, 5 figures

Multi-Access Edge Computing (MEC) is a key enabling technology for Fifth Generation (5G) mobile networks. MEC facilitates distributed cloud computing capabilities and information technology service environment for applications and services at the edges of mobile networks. This architectural modification serves to reduce congestion, latency, and improve the performance of such edge colocated applications and devices. In this paper, we demonstrate how reactive service migration can be orchestrated for low-power MEC-enabled Internet of Things (IoT) devices. Here, we use open-source Kubernetes as container orchestration system. Our demo is based on traditional client-server system from user equipment (UE) over Long Term Evolution (LTE) to the MEC server. As the use case scenario, we post-process live video received over web real-time communication (WebRTC). Next, we integrate orchestration by Kubernetes with S1 handovers, demonstrating MEC-based software defined network (SDN). Now, edge applications may reactively follow the UE within the radio access network (RAN), expediting low-latency. The collected data is used to analyze the benefits of the low-power MEC-enabled IoT device scheme, in which end-to-end (E2E) latency and power requirements of the UE are improved. We further discuss the challenges of implementing such schemes and future research directions therein.

Edge computing that leverages cloud resources to the proximity of user devices is seen as the future infrastructure for distributed applications. However, developing and deploying edge applications, that rely on cellular networks, is burdensome. Such network infrastructures are often based on proprietary components, each with unique programming abstractions and interfaces. To facilitate straightforward deployment of edge applications, we introduce OSS based RAN on OTA commercial spectrum with DevOps capabilities. OSS allows software modifications and integrations of the system components, e.g., EPC and edge hosts running applications, required for new data pipelines and optimizations not addressed in standardization. Such an OSS infrastructure enables further research and prototyping of novel end-user applications in an environment familiar to software engineers without telecommunications background. We evaluated the presented infrastructure with E2E OTA testing, resulting in 7.5MB/s throughput and latency of 21ms, which shows that the presented infrastructure provides low latency for edge applications.

This article analyzes the role of Finnish regulation in achieving the broadband penetration goals defined by the National Regulatory Authority. It is well known that in the absence of regulatory mitigation the population density has a positive effect on broadband diffusion. Hence, we measure the effect of the population density on the determinants of broadband diffusion throughout the postal codes of Finland via Geographically Weighted Regression. We suggest that the main determinants of broadband diffusion and the population density follow a spatial pattern that is either concentric with a weak/medium/strong strength or non-concentric convex/concave. Based on 10 patterns, we argue that the Finnish spectrum policy encouraged Mobile Network Operators to satisfy ambitious Universal Service Obligations without the need for a Universal Service Fund. Spectrum auctions facilitated infrastructure-based competition via equitable spectrum allocation and coverage obligation delivery via low-fee licenses. However, state subsidies for fiber deployment did not attract investment from nationwide operators due to mobile preference. These subsidies encouraged demand-driven investment, leading to the emergence of fiber consumer cooperatives. To explain this emergence, we show that when population density decreases, the level of mobile service quality decreases and community commitment increases. Hence, we recommend regulators implementing market-driven strategies for 5G to stimulate local investment. For example, by allocating the 3.5 GHz and higher bands partly through local light licensing., Comment: Improved version of the accepted manuscript by Telematics and Informatics

We have investigated surface shear waves at 22 MHz in a 0.5-micron-thick polymer film on SiO2/Si substrate at low temperatures using suspended and non-suspended graphene as detectors. By tracking ultrasound modes detected by oscillations of a trilayer graphene membrane both in vacuum and in helium superfluid, we assign the resonances to surface shear modes, generalized Love waves, in the resist/silicon-substrate system loaded with gold. The propagation velocity of these shear modes displays a logarithmic temperature dependence below 1 K, which is characteristic for modification of the elastic properties of a disordered solid owing to a large density of two level state (TLS) systems. For the dissipation of the shear mode, we find a striking logarithmic temperature dependence, which indicates a basic relation between the speed of the surface wave propagation and the mode dissipation., Comment: 16 pages, 5 figures

We propose here a scheme, based on the measurement of quadrature phase coherence, aimed at testing the Clauser-Horne-Shimony-Holt Bell inequality in an optomechanical setting. Our setup is constituted by two optical cavities dispersively coupled to a common mechanical resonator. We show that it is possible to generate EPR-like correlations between the quadratures of the output fields of the two cavities, and, depending on the system parameters, to observe the violation of the Clauser-Horne-Shimony-Holt inequality., Comment: 13 pages and 5 figures (typo fixed)

Based on the experimental evidence that impurities contribute to the dissipation properties of solid-state open quantum systems, we provide here a description in terms of nonlinear quantum Langevin equations of the role played by two-level systems in the dynamics of a bosonic degree of freedom. Our starting point is represented by the description of the system/environment coupling in terms of coupling to two separate reservoirs, modelling the interaction with external bosonic modes and two level systems, respectively. Furthermore, we show how this model represents a specific example of a class of open quantum systems that can be described by nonlinear quantum Langevin equations. Our analysis offers a potential explanation of the parametric effects recently observed in circuit-QED cavity optomechanics experiments., Comment: 10 pages, 5 figures

X-ray tomography is a reliable tool for determining the inner structure of 3D object with penetrating X-rays. However, traditional reconstruction methods such as FDK require dense angular sampling in the data acquisition phase leading to long measurement times, especially in X-ray micro-tomography to obtain high resolution scans. Acquiring less data using greater angular steps is an obvious way for speeding up the process and avoiding the need to save huge data sets available memory. However, computing 3D reconstruction from such a sparsely sampled dataset is very sensitive to measurement noise and modelling errors. An automatic regularization method is proposed for robust reconstruction, based on enforcing sparsity in the three-dimensional shearlet transform domain. The inputs of the algorithm are the projection data and {\it a priori} known expected degree of sparsity, denoted $0<{\mathcal C}_{pr}\leq 1$. The number ${\mathcal C}_{pr}$ can be calibrated from a few dense-angle reconstructions and fixed. Human subchondral bone samples were tested and morphometric parameters of the bone reconstructions were then analyzed using standard metrics. The proposed method is shown to outperform the baseline algorithm (FDK) in the case of sparsely collected data. The number of X-ray projections can be reduced up to 10\% of the total amount while retaining the quality of the reconstruction images and of the morphometric paramaters.

We investigate the CMB $\mu$ distortion in models where two uncorrelated sources contribute to primordial perturbations. We parameterise each source by an amplitude, tilt, running and running of the running. We perform a detailed analysis of the distribution signal as function of the model parameters, highlighting the differences compared to single-source models. As a specific example, we also investigate the mixed inflaton-curvaton scenario. We find that the $\mu$ distortion could efficiently break degeneracies of curvaton parameters especially when combined with future sensitivity of probing the tensor-to-scalar ratio $r$. For example, assuming bounds $\mu < 0.5 \times 10^{-8}$ and $r<0.01$, the curvaton contribution should either vanish or the curvaton should dominate primordial perturbations and its slow-roll parameter $\eta_{\chi}$ is constrained to the interval $-0.007 < \eta_{\chi}< 0.045$., Comment: 15 pages, 6 figures. v2: published version, minor changes

We report an accessible and robust tool for evaluating the effects of Coulomb collisions on a test particle in a plasma that obeys Maxwell-J\"uttner statistics. The implementation is based on the Beliaev-Budker collision integral which allows both the test particle and the background plasma to be relativistic. The integration method supports adaptive time stepping, which is shown to greatly improve the computational efficiency. The Monte Carlo method is implemented for both the three-dimensional particle momentum space and the five-dimensional guiding center phase space. Detailed description is provided for both the physics and implementation of the operator. The focus is in adaptive integration of stochastic differential equations, which is an overlooked aspect among existing Monte Carlo implementations of Coulomb collision operators. We verify that our operator converges to known analytical results and demonstrate that careless implementation of the adaptive time step can lead to severely erroneous results. The operator is provided as a self-contained Fortran 95 module and can be included into existing orbit-following tools that trace either the full Larmor motion or the guiding center dynamics. The adaptive time-stepping algorithm is expected to be useful in situations where the collision frequencies vary greatly over the course of a simulation. Examples include the slowing-down of fusion products or other fast ions, and the Dreicer generation of runaway electrons as well as the generation of fast ions or electrons with ion or electron cyclotron resonance heating.