Knowledge distillation methods are proved to be promising in improving the performance of neural networks and no additional computational expenses are required during the inference time. For the sake of boosting the accuracy of object detection, a great number of knowledge distillation methods have been proposed particularly designed for object detection. However, most of these methods only focus on feature-level distillation and label-level distillation, leaving the label assignment step, a unique and paramount procedure for object detection, by the wayside. In this work, we come up with a simple but effective knowledge distillation approach focusing on label assignment in object detection, in which the positive and negative samples of student network are selected in accordance with the predictions of teacher network. Our method shows encouraging results on the MSCOCO2017 benchmark, and can not only be applied to both one-stage detectors and two-stage detectors but also be utilized orthogonally with other knowledge distillation methods.
In this paper we study a singular limit problem in the context of partially dissipative first order quasilinear systems. This problem arises in multiphase fluid mechanics. More precisely, taking into account dissipative effects for the velocity, we show that the so-called Kapilla system is obtained as a relaxation limit from the Baer-Nunziato (BN) system and derive the convergence rate of this process. The main problem we encounter is that the (BN)-system does not verify the celebrated (SK) condition due to Shizuta and Kawashima. It turns out that we can rewrite the (BN)-system in terms of new variables such as to highlight a subsystem for which the linearized does verify the (SK) condition which is coupled through lower-order terms with a transport equation. We construct an appropriate weighted energy-functional which allows us to tackle the lack of symmetry of the system, provides decay information and allows us to close the estimate uniformly with respect to the relaxation parameter.
We study high order harmonics generation (HHG) in crystalline silicon and diamond subjected to near and mid-infrared laser pulses. We employ time-dependent density functional theory and solve the time-dependent Kohn-Sham equation in the single-cell geometry. We demonstrate that clear and clean HHG spectra can be generated with careful selection of the pulse duration. In addition, we implement dephasing effects through a displacement of atomic positions in a silicon large super-cell prepared by a molecular dynamics simulation. We compare our results with the previous calculations by Floss et al. [arXiv:1705.10707] [Phys. Rev. A 97, 011401(R) (2018)] on Diamond at 800 nm and by Tancogne-Dejean et al. [arXiv:1609.09298] [Phys. Rev. Lett. 118, 087403 (2017)] on Si at 2000 nm.
Context. Insight into the conditions that drive the physics and chemistry in interstellar clouds is gained from determining the abundance and charge state of their components. Aims. We propose an evaluation of the C60:C60+ ratio in diffuse and translucent interstellar clouds that exploits electronic absorption bands so as not to rely on ambiguous IR emission measurements. Methods. The ratio is determined by analyzing archival spectra and literature data. Information on the cation population is obtained from published characteristics of the main diffuse interstellar bands attributed to C60+ and absorption cross sections already reported for the vibronic bands of the cation. The population of neutral molecules is described in terms of upper limit because the relevant vibronic bands of C60 are not brought out by observations. We revise the oscillator strengths reported for C60 and measure the spectrum of the molecule isolated in Ne ice to complete them. Results. We scale down the oscillator strengths for absorption bands of C60 and find an upper limit of approximately 1.3 for the C60:C60+ ratio. Conclusions. We conclude that the fraction of neutral molecules in the buckminsterfullerene population of diffuse and translucent interstellar clouds may be notable despite the non-detection of the expected vibronic bands. More certainty will require improved laboratory data and observations.
Neutron diffraction studies of HoFeO$_3$ single crystal were performed under external magnetic fields. The interplay between the external magnetic field, Dzyaloshinsky-Moria antisymmetric exchange and isotropic exchange interactions between Fe and Ho sublattice and inside Fe sublattice provides a rich phase diagram. As the result of the balance of exchange interactions inside crystal and external magnetic field we found 8 different magnetic phases, produced or suppressed by the field.
The LISA mission will observe gravitational waves emitted from tens of thousands of galactic binaries, in particular white dwarf binary systems. These objects are known to have intense magnetic fields. However, these fields are usually not considered as their influence on the orbital and rotational motion of the binary is assumed for being too weak. It turns out that magnetic fields modify the orbits, in particular their geometry with respect to the observer. In this work, we revisit the issue, assuming magnetostatic approximation, and we show how the magnetic fields within a binary system generate a secular drift in the argument of the periastron, leading then, to modifications of the gravitational waveforms that are potentially detectable by LISA.
Artificial Intelligence (AI) is rapidly becoming integrated into military Command and Control (C2) systems as a strategic priority for many defence forces. The successful implementation of AI is promising to herald a significant leap in C2 agility through automation. However, realistic expectations need to be set on what AI can achieve in the foreseeable future. This paper will argue that AI could lead to a fragility trap, whereby the delegation of C2 functions to an AI could increase the fragility of C2, resulting in catastrophic strategic failures. This calls for a new framework for AI in C2 to avoid this trap. We will argue that antifragility along with agility should form the core design principles for AI-enabled C2 systems. This duality is termed Agile, Antifragile, AI-Enabled Command and Control (A3IC2). An A3IC2 system continuously improves its capacity to perform in the face of shocks and surprises through overcompensation from feedback during the C2 decision-making cycle. An A3IC2 system will not only be able to survive within a complex operational environment, it will also thrive, benefiting from the inevitable shocks and volatility of war.
Full 3D scanning can efficiently be obtained using structured light scanning combined with a rotation stage. In this setting it is, however, necessary to reposition the object and scan it in different poses in order to cover the entire object. In this case, correspondence between the scans is lost, since the object was moved. In this paper, we propose a fully automatic method for aligning the scans of an object in two different poses. This is done by matching 2D features between images from two poses and utilizing correspondence between the images and the scanned point clouds. To demonstrate the approach, we present the results of scanning three dissimilar objects.
We discuss diagonalization of propagator for mixing fermions system based on the eigenvalue problem. The similarity transformation converting matrix propagator into diagonal form is obtained. The suggested diagonalization has simple algebraic properties for on-shell fermions and can be used in renormalization of fermion mixing matrix.
Penetration testing the organised attack of a computer system in order to test existing defences has been used extensively to evaluate network security. This is a time consuming process and requires in-depth knowledge for the establishment of a strategy that resembles a real cyber-attack. This paper presents a novel deep reinforcement learning architecture with hierarchically structured agents called HA-DRL, which employs an algebraic action decomposition strategy to address the large discrete action space of an autonomous penetration testing simulator where the number of actions is exponentially increased with the complexity of the designed cybersecurity network. The proposed architecture is shown to find the optimal attacking policy faster and more stably than a conventional deep Q-learning agent which is commonly used as a method to apply artificial intelligence in automatic penetration testing.
