Search

Existing learning rate schedules that do not require specification of the optimization stopping step T are greatly out-performed by learning rate schedules that depend on T. We propose an approach that avoids the need for this stopping time by eschewing the use of schedules entirely, while exhibiting state-of-the-art performance compared to schedules across a wide family of problems ranging from convex problems to large-scale deep learning problems. Our Schedule-Free approach introduces no additional hyper-parameters over standard optimizers with momentum. Our method is a direct consequence of a new theory we develop that unifies scheduling and iterate averaging. An open source implementation of our method is available at https://github.com/facebookresearch/schedule_free. Schedule-Free AdamW is the core algorithm behind our winning entry to the MLCommons 2024 AlgoPerf Algorithmic Efficiency Challenge Self-Tuning track.

Skyrmions are topological structures characterized by a winding vectorial configuration that provides a quantized topological charge. In magnetic materials, skyrmions are localized spin textures that exhibit unique stability and mobility properties, making them highly relevant to the burgeoning field of spintronics. In optics, these structures open new frontiers in manipulating and controlling light at the nanoscale. The convergence of optics and magnetics holds therefore immense potential for manipulating magnetic processes at ultrafast timescales. Here, we explore the possibility of generating skyrmionic topological structures within the magnetic field induced by the inverse Faraday effect in a plasmonic nanostructure. Our investigation reveals that a gold nanoring, featuring a dark mode, can generate counter-propagating photocurrents between its inner and outer segments, thereby enabling the magnetization of gold and supporting a skyrmionic vectorial distribution. We elucidate that these photocurrents arise from the localized control of light polarization, facilitating their counter-propagative motion. The generation of skyrmions through the inverse Faraday effect at the nanoscale presents a pathway towards directly integrating this topology into magnetic layers. This advancement holds promise for ultrafast timescales, offering direct applications in ultrafast data writing and processing.

Photocurrents play a crucial role in various applications, including light detection, photovoltaics, and THz radiation generation. Despite the abundance of methods and materials for converting light into electrical signals, the use of metals in this context has been relatively limited. Nanostructures supporting surface plasmons in metals offer precise light manipulation and induce light-driven electron motion. Through inverse design optimization of a gold nanostructure, we demonstrate enhanced volumetric, unidirectional, intense, and ultrafast photocurrents via a magneto-optical process derived from the inverse Faraday effect. This is achieved through fine-tuning the amplitude, polarization, and their gradients in the local light field. The virtually instantaneous process allows dynamic photocurrent modulation by varying optical pulse duration, potentially yielding nanosources of intense, ultrafast, planar magnetic fields, and frequency-tunable THz emission. These findings opens avenues for ultrafast magnetic material manipulation and holds promise for nanoscale THz spectroscopy., Comment: 5 figures

Hand-eye calibration, as a fundamental task in vision-based robotic systems, aims to estimate the transformation matrix between the coordinate frame of the camera and the robot flange. Most approaches to hand-eye calibration rely on external markers or human assistance. We proposed Look at Robot Base Once (LRBO), a novel methodology that addresses the hand-eye calibration problem without external calibration objects or human support, but with the robot base. Using point clouds of the robot base, a transformation matrix from the coordinate frame of the camera to the robot base is established as I=AXB. To this end, we exploit learning-based 3D detection and registration algorithms to estimate the location and orientation of the robot base. The robustness and accuracy of the method are quantified by ground-truth-based evaluation, and the accuracy result is compared with other 3D vision-based calibration methods. To assess the feasibility of our methodology, we carried out experiments utilizing a low-cost structured light scanner across varying joint configurations and groups of experiments. The proposed hand-eye calibration method achieved a translation deviation of 0.930 mm and a rotation deviation of 0.265 degrees according to the experimental results. Additionally, the 3D reconstruction experiments demonstrated a rotation error of 0.994 degrees and a position error of 1.697 mm. Moreover, our method offers the potential to be completed in 1 second, which is the fastest compared to other 3D hand-eye calibration methods. Code is released at github.com/leihui6/LRBO., Comment: Accepted by Journal of Intelligent & Robotic Systems

Composed image retrieval is a type of image retrieval task where the user provides a reference image as a starting point and specifies a text on how to shift from the starting point to the desired target image. However, most existing methods focus on the composition learning of text and reference images and oversimplify the text as a description, neglecting the inherent structure and the user's shifting intention of the texts. As a result, these methods typically take shortcuts that disregard the visual cue of the reference images. To address this issue, we reconsider the text as instructions and propose a Semantic Shift network (SSN) that explicitly decomposes the semantic shifts into two steps: from the reference image to the visual prototype and from the visual prototype to the target image. Specifically, SSN explicitly decomposes the instructions into two components: degradation and upgradation, where the degradation is used to picture the visual prototype from the reference image, while the upgradation is used to enrich the visual prototype into the final representations to retrieve the desired target image. The experimental results show that the proposed SSN demonstrates a significant improvement of 5.42% and 1.37% on the CIRR and FashionIQ datasets, respectively, and establishes a new state-of-the-art performance. Codes will be publicly available.

Seismic data noise processing is an important part of seismic exploration data processing, and the effect of noise elimination is directly related to the follow-up processing of data. In response to this problem, many authors have proposed methods based on rank reduction, sparse transformation, domain transformation, and deep learning. However, such methods are often not ideal when faced with strong noise. Therefore, we propose to use diffusion model theory for noise removal. The Bayesian equation is used to reverse the noise addition process, and the noise reduction work is divided into multiple steps to effectively deal with high-noise situations. Furthermore, we propose to evaluate the noise level of blind Gaussian seismic data using principal component analysis to determine the number of steps for noise reduction processing of seismic data. We train the model on synthetic data and validate it on field data through transfer learning. Experiments show that our proposed method can identify most of the noise with less signal leakage. This has positive significance for high-precision seismic exploration and future seismic data signal processing research., Comment: 10 pages, 13 figures. This work has been submitted to the IEEE for possible publication

The inverse Faraday effect is a magneto-optical process allowing the magnetization of matter by an optical excitation carrying a non-zero spin of light. In particular, a right circular polarization generates a magnetization in the direction of light propagation and a left circular polarization in the opposite direction to this propagation. We demonstrate here that by manipulating the spin density of light, i.e., its polarization, in a plasmonic nanostructure, we generate a reversed inverse Faraday effect. A right circular polarization will generate a magnetization in the opposite direction of the light propagation, a left circular polarization in the direction of propagation. Also, we demonstrate that this new physical phenomenon is chiral, generating a strong magnetic field only for one helicity of the light, the opposite helicity producing this effect only for the mirror structure. This new optical concept opens the way to the generation of magnetic fields with unpolarized light, finding application in the ultrafast manipulation of magnetic domains and processes, such as spin precession, spin currents, and waves, magnetic skyrmion or magnetic circular dichroism, with direct applications in data storage and processing technologies., Comment: arXiv admin note: text overlap with arXiv:2301.05971

Magnetic monopoles are hypothetical particles that, like electric monopoles which generate electric fields, are at the origin of magnetic fields. Despite many efforts, to date, these theoretical particles have yet to be observed. Nevertheless, many systems or physical phenomena can be related to magnetic monopole behavior. Here, we propose a new type of photonic nanoantenna behaving as a radiating magnetic monopole. We demonstrate that a half-nanoslit in a semi-infinite gold layer generates a single pole of an enhanced magnetic field at the nanoscale and that this single pole radiates efficiently in the far field. This original antenna concept opens the way to a new model system to study magnetic monopoles, to a new source of optical magnetic field to study the "magnetic light" and matter coupling, and allows potential applications at other frequencies such as magnetic resonance imaging.

Sensitivity to molecular chirality is crucial for many fields, from biology and chemistry to the pharmaceutical industry. By generating superchiral light, nanophotonics has brought innovative solutions to reduce the detection volume and increase sensitivity at the cost of a non-selectivity of light chirality or a strong contribution to the background. Here, we theoretically propose an achiral plasmonic resonator, based on a rectangular nanoslit in a thin gold layer behaving as a magnetic dipole, to generate a tunable nanosource of purely superchiral light. This nanosource is free of any background, and the sign of its chirality is externally tunable in wavelength and polarization. These properties result from the coupling between the incident wave and the magnetic dipolar character of our nano-antenna. Thus, our results propose a platform with deep subwavelength detection volumes for chiral molecules in particular, in the visible, and a roadmap for optimizing the signal-to-noise ratios in circular dichroism measurements to reach single-molecule sensitivity.

Light-matter interactions are often considered governed by the electric optical field only, leaving aside the magnetic component of light. However, the magnetic part plays a determining role in many optical processes from light and chiral-matter interactions, photon-avalanching to forbidden photochemistry, making the manipulation of magnetic processes extremely relevant. Here, by creating a standing wave using a plasmonic nanomirror we manipulate the spatial distributions of the electric and magnetic fields and their associated local density of states, allowing the selective control of the excitation and emission of electric and magnetic dipolar transitions. This control allows us to image, in 3D, the electric and magnetic nodes and anti-nodes of the fields interference pattern. It also enables us to enhance specifically photoluminescence from quantum emitters excited only by the magnetic field, and to manipulate their quantum environment by acting on the excitation fields solely, demonstrating full control of magnetic and electric light-matter interactions.

The inverse Faraday effect is a magneto-optical process allowing the magnetization of matter by an optical excitation carrying a non-zero spin or orbital moment of light. This phenomenon was considered until now as symmetric; right or left circular polarizations generate magnetic fields oriented in the direction of light propagation or in the counter-propagating direction. Here, we demonstrate that by manipulating the spin density of light in a plasmonic nanostructure, we generate a chiral inverse Faraday effect, creating a strong magnetic field of 500 mT only for one helicity of the light, the opposite helicity producing this effect only for the mirror structure. This new optical concept opens the way to the generation of magnetic fields with unpolarized light, finding application in the ultrafast manipulation of magnetic domains and processes, such as spin precession, spin currents and waves, magnetic skyrmion or magnetic circular dichroism, with direct applications in data storage and data processing technologies.

The inverse Faraday effect (IFE) allows the generation of magnetic fields by optical excitation only. Since its discovery in the 60s, it was believed that only an elliptical or circular polarization could magnetize matter by this magneto-optical phenomenon. Here, we demonstrate the generation of an IFE via a linear polarization of light. This new physical concept results from the local manipulation of light by a plasmonic nano-antenna. We demonstrate that a gold nanorod excited by a linear polarization generates a non-zero magnetic field by IFE when the incident polarization of the light is not parallel to the long axis of the rod. We show that this dissymmetry generates hot spots of local non-vanishing spin densities (local elliptical polarization state), introducing the concept of super circular light, allowing this magnetization. Moreover, by varying the angle of the incident linear polarization with respect to the nano-antenna, we demonstrate the on-demand flipping of the magnetic field orientation. Finally, this linear IFE generates a stationary magnetic field 25 times stronger than what a gold nanoparticle produces when excited by a circular polarization and via a classical IFE. The creation of stationary magnetic fields by IFE in a plasmonic nanostructure is nowadays the only technique allowing the creation of ultra-short, intense magnetic field pulses at the nanoscale. Thus, it finds applications in the ultrafast control of magnetic domains with applications not only in data storage technologies but also in research fields such as magnetic trapping, magnetic skyrmion, magnetic circular dichroism, to spin control, spin precession, spin currents, and spin waves, among others., Comment: 13 pages, 6 figures, article