Search

In deep reinforcement learning applications, maximizing discounted reward is often employed instead of maximizing total reward to ensure the convergence and stability of algorithms, even though the performance metric for evaluating the policy remains the total reward. However, the optimal policies corresponding to these two objectives may not always be consistent. To address this issue, we analyzed the suboptimality of the policy obtained through maximizing discounted reward in relation to the policy that maximizes total reward and identified the influence of hyperparameters. Additionally, we proposed sufficient conditions for aligning the optimal policies of these two objectives under various settings. The primary contributions are as follows: We theoretically analyzed the factors influencing performance when using discounted reward as a proxy for total reward, thereby enhancing the theoretical understanding of this scenario. Furthermore, we developed methods to align the optimal policies of the two objectives in certain situations, which can improve the performance of reinforcement learning algorithms.

Semi-gradient Q-learning is applied in many fields, but due to the absence of an explicit loss function, studying its dynamics and implicit bias in the parameter space is challenging. This paper introduces the Fokker--Planck equation and employs partial data obtained through sampling to construct and visualize the effective loss landscape within a two-dimensional parameter space. This visualization reveals how the global minima in the loss landscape can transform into saddle points in the effective loss landscape, as well as the implicit bias of the semi-gradient method. Additionally, we demonstrate that saddle points, originating from the global minima in loss landscape, still exist in the effective loss landscape under high-dimensional parameter spaces and neural network settings. This paper develop a novel approach for probing implicit bias in semi-gradient Q-learning.

Quantum heat engines and refrigerators are open quantum systems, whose dynamics can be well understood using a non-Hermitian formalism. A prominent feature of non-Hermiticity is the existence of exceptional points (EPs), which has no counterpart in closed quantum systems. It has been shown in classical systems that dynamical encirclement in the vicinity of an EP, whether the loop includes the EP or not, could lead to chiral mode conversion. Here, we show that this is valid also for quantum systems when dynamical encircling is performed in the vicinity of their Liouvillian EPs (LEPs) which include the effects of quantum jumps and associated noise - an important quantum feature not present in previous works. We demonstrate, using a Paul-trapped ultracold ion, the first chiral quantum heating and refrigeration by dynamically encircling a closed loop in the vicinity of an LEP. We witness the cycling direction to be associated with the chirality and heat release (absorption) of the quantum heat engine (quantum refrigerator). Our experiments have revealed that not only the adiabaticity-breakdown but also the Landau-Zener-St\"uckelberg process play an essential role during dynamic encircling, resulting in chiral thermodynamic cycles. Our observations contributes to further understanding of chiral and topological features in non-Hermitian systems and pave a way to exploring the relation between chirality and quantum thermodynamics., Comment: Accepted by Light: Science & Applications

Task-oriented object detection aims to find objects suitable for accomplishing specific tasks. As a challenging task, it requires simultaneous visual data processing and reasoning under ambiguous semantics. Recent solutions are mainly all-in-one models. However, the object detection backbones are pre-trained without text supervision. Thus, to incorporate task requirements, their intricate models undergo extensive learning on a highly imbalanced and scarce dataset, resulting in capped performance, laborious training, and poor generalizability. In contrast, we propose TaskCLIP, a more natural two-stage design composed of general object detection and task-guided object selection. Particularly for the latter, we resort to the recently successful large Vision-Language Models (VLMs) as our backbone, which provides rich semantic knowledge and a uniform embedding space for images and texts. Nevertheless, the naive application of VLMs leads to sub-optimal quality, due to the misalignment between embeddings of object images and their visual attributes, which are mainly adjective phrases. To this end, we design a transformer-based aligner after the pre-trained VLMs to re-calibrate both embeddings. Finally, we employ a trainable score function to post-process the VLM matching results for object selection. Experimental results demonstrate that our TaskCLIP outperforms the state-of-the-art DETR-based model TOIST by 3.5% and only requires a single NVIDIA RTX 4090 for both training and inference.

Very recently, the Belle~II Collaboration presented a measurement for the decays $B^+\to\bar{D}^{(*)0} K^+\bar{K}^0$ and $B^0\to D^{(*)-}K^+\bar{K}^0$, the bulk of observed $m(K^+ K_S^0)$ distributions showing low-mass structures in all four channels. In this work, we study the contributions of $\rho(770,1450)^+$, $a_2(1320)^+$ and $a_0(980,1450)^+$ resonances to these decay processes. The intermediate states $\rho(770,1450)^+$ are found to dominate the low-mass distribution of kaon pairs roughly contributing to half of the total branching fraction in each of the four decay channels. The contribution of the tensor $a_2(1320)^+$ meson is found to be negligible. Near the threshold of the kaon pair, the state $a_0(980)^+$ turns out to be much less important than expected, not being able to account for the enhancement of events in that energy region observed in the $B^+\to\bar{D}^{(*)0} K^+\bar{K}^0$ decays. Further studies both from the theoretical and experimental sides are needed to elucidate the role of the non-resonant contributions governing the formation of $K^+\bar{K}^0$ pairs near their threshold in these decay processes., Comment: 18 pages, 4 figures

In recent times, a plethora of hardware accelerators have been put forth for graph learning applications such as vertex classification and graph classification. However, previous works have paid little attention to Knowledge Graph Completion (KGC), a task that is well-known for its significantly higher algorithm complexity. The state-of-the-art KGC solutions based on graph convolution neural network (GCN) involve extensive vertex/relation embedding updates and complicated score functions, which are inherently cumbersome for acceleration. As a result, existing accelerator designs are no longer optimal, and a novel algorithm-hardware co-design for KG reasoning is needed. Recently, brain-inspired HyperDimensional Computing (HDC) has been introduced as a promising solution for lightweight machine learning, particularly for graph learning applications. In this paper, we leverage HDC for an intrinsically more efficient and acceleration-friendly KGC algorithm. We also co-design an acceleration framework named HDReason targeting FPGA platforms. On the algorithm level, HDReason achieves a balance between high reasoning accuracy, strong model interpretability, and less computation complexity. In terms of architecture, HDReason offers reconfigurability, high training throughput, and low energy consumption. When compared with NVIDIA RTX 4090 GPU, the proposed accelerator achieves an average 10.6x speedup and 65x energy efficiency improvement. When conducting cross-models and cross-platforms comparison, HDReason yields an average 4.2x higher performance and 3.4x better energy efficiency with similar accuracy versus the state-of-the-art FPGA-based GCN training platform.

Deep reinforcement learning excels in numerous large-scale practical applications. However, existing performance analyses ignores the unique characteristics of continuous-time control problems, is unable to directly estimate the generalization error of the Bellman optimal loss and require a boundedness assumption. Our work focuses on continuous-time control problems and proposes a method that is applicable to all such problems where the transition function satisfies semi-group and Lipschitz properties. Under this method, we can directly analyze the \emph{a priori} generalization error of the Bellman optimal loss. The core of this method lies in two transformations of the loss function. To complete the transformation, we propose a decomposition method for the maximum operator. Additionally, this analysis method does not require a boundedness assumption. Finally, we obtain an \emph{a priori} generalization error without the curse of dimensionality.

We study the contributions for $K\eta^{\prime}$ pair originating from the scalar intermediate state $K_0^{*}(1430)$ in the three-body decays $B\to K\eta^{\prime} h$ ($h=\pi, K$) within the perturbative QCD approach. The contribution of $K^*_0(1430)\to K\eta^{\prime}$ is described by the Flatt${\rm \acute{e}}$ formula with coupled channels $K\pi$, $K\eta$ and $K\eta^{\prime}$. The strong coupling constants $g_{K^*_0K\eta^{(\prime)}}$ are extracted from $g_{K^*_0 K\pi}$ within flavor SU$(3)$ symmetry. In spite of the strong depression by phase space near the threshold of $K\eta^\prime$, the $CP$ averaged branching fractions for the $B\to K^*_0(1430) h \to K\eta^\prime h$ decays are predicted to be on the order of $10^{-8}$ to $10^{-5}$, which are non-negligible for the corresponding three-body $B$ decays. Since the K$\eta$ system is almost decoupled from the even-spin strange mesons under flavor SU$(3)$ symmetry, those quasi-two-body $B$ decays with subprocess $K^*_0(1430) \to K \eta$ shall have quite small branching ratios and are not taken into account in this work. We also estimate that the branching fraction for $K_0^{*}(1430)\to K\eta^{\prime}$ is about one fifth of that for $K_0^{*}(1430)\to K\pi$. The predictions for the relevant decays are expected to be tested by the LHCb and Belle-II experiments in the future., Comment: 10 pages, 2 figures and 4 tables. Matching the published version in PRD

The scanning tunneling microscopy induced luminescence (STML) can be used to probe the optical and electronic properties of molecules. Concerning the vibronic coupling, we model the molecule as a two-level system with the vibrational degrees of freedom. Based on the Bardeen's theory, we express the inelastic tunneling current in terms of Huang-Rhys factor within the inelastic electron scattering (IES) mechanism. We find that the differential conductance, varying with the bias voltage, exhibits distinct step structure with various vibronic coupling strength. The second derivative of the inelastic tunneling current with respect to the bias voltage shows the characteristics of vibrational-level structure with Franck-Condon factor. Consequently, we propose a method to determine the Huang-Rhys factor of molecules, holding promising potential within the realm of solid-state physics., Comment: 16 pages, 3 figures

The decays $B\to\bar{D}^{(*)} \omega\pi$ are very important for the investigation of $\rho$ excitations and the test of factorization hypothesis for $B$ meson decays. The $B^{+}\to \bar{D}^{(*)0}\omega\pi^+$ and $B^{0}\to D^{(*)-}\omega\pi^+$ have been measured by different collaborations but without any predictions for their observables on theoretical side. In this work, we study the contributions of $\rho(770,1450)\to \omega\pi$ for the cascade decays $B^{+}\to \bar{D}^{(*)0} \rho^+ \to \bar{D}^{(*)0}\omega\pi^+$, $B^{0}\to D^{(*)-} \rho^+ \to D^{(*)-}\omega\pi^+$ and $B_s^{0}\to D_s^{(*)-} \rho^+ \to D^{(*)-}\omega\pi^+$. We introduce $\rho(770,1450)\to \omega\pi$ subprocesses into the distribution amplitudes for $\omega\pi$ system via the vector form factor $F_{\omega\pi}(s)$ and then predict the branching fractions for the first time for concerned quasi-two-body decays with $\rho(770,1450)\to \omega\pi$, as well as the corresponding longitudinal polarization fractions $\Gamma_L/\Gamma$ for the cases with the vector $\bar{D}^{*0}$ or $D_{(s)}^{*-}$ in their final states. The branching fractions of these quasi-two-body decays are predicted at the order of $10^{-3}$, which can be detected at the LHCb and Belle-II experiments. The predictions for the decays ${B}^0 \to{D}^{*-} \rho(770)^+\to {D}^{*-} \omega\pi^+$ and ${B}^0 \to {D}^{*-} \rho(1450)^+\to {D}^{*-} \omega\pi^+$ agree well with the measurements from Belle Collaboration. In order to avoid the pollution from annihilation Feynman diagrams, we recommend to take the $B_s^0 \to D_s^{*-}\rho(770,1450)^+$ decays, which have only emission diagrams at quark level, to test the factorization hypothesis for $B$ decays., Comment: 30 pages, 4 figures, the typos in Eq.(2.33) were corrected

This paper considers the joint compression and enhancement problem for speech signal in the presence of noise. Recently, the SoundStream codec, which relies on end-to-end joint training of an encoder-decoder pair and a residual vector quantizer by a combination of adversarial and reconstruction losses,has shown very promising performance, especially in subjective perception quality. In this work, we provide a theoretical result to show that, to simultaneously achieve low distortion and high perception in the presence of noise, there exist an optimal two-stage optimization procedure for the joint compression and enhancement problem. This procedure firstly optimizes an encoder-decoder pair using only distortion loss and then fixes the encoder to optimize a perceptual decoder using perception loss. Based on this result, we construct a two-stage training framework for joint compression and enhancement of noisy speech signal. Unlike existing training methods which are heuristic, the proposed two-stage training method has a theoretical foundation. Finally, experimental results for various noise and bit-rate conditions are provided. The results demonstrate that a codec trained by the proposed framework can outperform SoundStream and other representative codecs in terms of both objective and subjective evaluation metrics. Code is available at \textit{https://github.com/jscscloris/SEStream}.

Recent studies show that, without any prior model, the unsupervised restoration learning problem can be optimally formulated as an optimal transport (OT) problem, which has shown promising performance on denoising tasks to approach the performance of supervised methods. However, it still significantly lags behind state-of-the-art supervised methods on complex restoration tasks such as super-resolution, deraining, and dehazing. In this paper, we exploit the sparsity of degradation in the OT framework to significantly boost its performance on these tasks. First, we disclose an observation that the degradation in these tasks is quite sparse in the frequency domain, and then propose a sparsity-aware optimal transport (SOT) criterion for unsupervised restoration learning. Further, we provide an analytic example to illustrate that exploiting the sparsity helps to reduce the ambiguity in finding an inverse map for restoration. Experiments on real-world super-resolution, deraining, and dehazing demonstrate that SOT can improve the PSNR of OT by about 2.6 dB, 2.7 dB and 1.3 dB, respectively, while achieving the best perception scores among the compared supervised and unsupervised methods. Particularly, on the three tasks, SOT significantly outperforms existing unsupervised methods and approaches the performance of state-of-the-art supervised methods., Comment: 15 pages, 9 figures

Visual simultaneous localization and mapping (SLAM) systems face challenges in detecting loop closure under the circumstance of large viewpoint changes. In this paper, we present an object-based loop closure detection method based on the spatial layout and semanic consistency of the 3D scene graph. Firstly, we propose an object-level data association approach based on the semantic information from semantic labels, intersection over union (IoU), object color, and object embedding. Subsequently, multi-view bundle adjustment with the associated objects is utilized to jointly optimize the poses of objects and cameras. We represent the refined objects as a 3D spatial graph with semantics and topology. Then, we propose a graph matching approach to select correspondence objects based on the structure layout and semantic property similarity of vertices' neighbors. Finally, we jointly optimize camera trajectories and object poses in an object-level pose graph optimization, which results in a globally consistent map. Experimental results demonstrate that our proposed data association approach can construct more accurate 3D semantic maps, and our loop closure method is more robust than point-based and object-based methods in circumstances with large viewpoint changes.

The search for extraterrestrial intelligence (SETI) is to search for technosignatures associated with extraterrestrial life, such as engineered radio signals. In this paper, we apply the multibeam coincidence matching (MBCM) strategy, and propose a new search mode based on the MBCM which we call MBCM blind search mode. In our recent targeted SETI research, 33 exoplanet systems are observed by the Five-hundred-meter Aperture Spherical radio Telescope (FAST). With this blind search mode, we search for narrowband drifting signals across $1.05-1.45$ GHz in two orthogonal linear polarization directions separately. There are two special signals, one of which can only be detected by the blind search mode while the other can be found by both blind and targeted search modes. This result reveals huge advantages of the new blind search mode. However, we eliminate the possibility of the special signals being ETI signals based on much evidence, such as the polarization, drift, frequency and beam coverage characteristics. Our observations achieve an unprecedented sensitivity and our work provides a deeper understanding to the polarization analysis of extraterrestrial signals., Comment: 15 pages, 1 table, 7 figures. Matches AJ published version. Based on MBCM strategy by Zhen-Zhao Tao et al (2022 AJ 164 160), we propose a new search mode: MBCM blind search mode for FAST SETI observations

We study the $J/\psi \to \phi K \bar K$ decay, looking for differences in the production rates of $K^+ K^-$ or $K^0 \bar K^0$ in the region of 1700-1800 MeV, where two resonances appear dynamically generated from the vector-vector interaction. Two resonances are known experimentally in that region, the $f_0(1710)$ and a new resonance reported by the BABAR and BESIII collaborations. The $K \bar K$ should be produced with $I=0$ in that reaction, but due to the different $K^{*0}$ and $K^{*+}$ masses some isospin violation appears. Yet, due to the large width of the $K^*$, the violation obtained is very small and the rates of $K^+ K^-$ or $K^0 \bar K^0$ production are equal within $5\%$. However, we also find that due to the step needed to convert two vectors into $K \bar K$, a shape can appear in the $K \bar K$ mass distribution that can mimic the $a_0$ production around the $K^* \bar K^*$ threshold, and is simply a threshold effect., Comment: 13 pages; 11 figures

Fast radio bursts (FRBs) are highly dispersed millisecond-duration radio bursts, of which the physical origin is still not fully understood. FRB 20201124A is one of the most actively repeating FRBs. In this paper, we present the collection of 1863 burst dynamic spectra of FRB 20201124A measured with the Five-hundred-meter Aperture Spherical radio Telescope (FAST). The current collection, taken from the observation during the FRB active phase from April to June 2021, is the largest burst sample detected in any FRB so far. The standard PSRFITs format is adopted, including dynamic spectra of the burst, and the time information of the dynamic spectra, in addition, mask files help readers to identify the pulse positions are also provided.

We have a look at the $P_{cs}$ states generated from the interaction of $\bar D^{(*)} \Xi^{(\prime*)}_c$ coupled channels. We consider the blocks of pseudoscalar-baryon $({\frac12}^+, {\frac32}^+)$ and vector-baryon $({\frac12}^+, {\frac32}^+)$, and find $10$ resonant states coupling mostly to $\bar D \Xi_c, \bar D^* \Xi_c,\bar D \Xi'_c, \bar D^* \Xi'_c,\bar D \Xi^*_c$ and $\bar D^* \Xi^*_c$. A novel aspect of the work is the realization that the $\bar D \Xi_c,\bar D_s\Lambda_c$ or $\bar D^* \Xi_c,\bar D^*_s\Lambda_c$ channels, with a strong transition potential, collaborate to produce a larger attraction than the corresponding states $\bar D \Sigma_c,\bar D\Lambda_c$ or $\bar D^* \Sigma_c,\bar D^*\Lambda_c$ appearing in the generation of the strangenessless $P_{c}$ states, since in the latter case the transition potential between those channels is zero. The extra attraction obtained in the $\bar D \Xi_c,\bar D^* \Xi_c$ pairs preclude the association of these channels to the $P_{cs}(4338)$ and $P_{cs}(4459)$ states respectively. Then we find a natural association of the $P_{cs}(4338)$ state coupling mostly to $\bar D^* \Xi_c$ while the $P_{cs}(4459)$ is associated to the state found that couples mostly to $\bar D \Xi'_c$. Four more states appear, like in other molecular pictures, and some of the states are degenerate in spin. Counting different spin states we find $10$ states, which we hope can be observed in the near future., Comment: 8 pages, 4 figures, accepted for publication in Physics Letters B

We study the interaction of meson-baryon coupled channels carrying quantum numbers of $\Omega_{cc}$ , $\Omega_{bb}$ and $\Omega_{bc}$ presently under investigation by the LHCb collaboration. The interaction is obtained from an extension of the local hidden gauge approach to the heavy quark sector that has proved to provide accurate results compared to experiment in the case of $\Omega_{c}$ , $\Xi_{c}$ states and pentaquarks, $P_c$ and $P_{cs}$. We obtain many bound states, with small decay widths within the space of the chosen coupled channels. The spin-parity of the states are $J^P={\frac{1}{2}}^-$ for coupled channels of pseudoscalar-baryon (${\frac{1}{2}}^+$), $J^P={\frac{3}{2}}^-$ for the case of pseudoscalar-baryon (${\frac{3}{2}}^+$), $J^P={\frac{1}{2}}^-,{\frac{3}{2}}^-$ for the case of vector-baryon (${\frac{1}{2}}^+$) and $J^P={\frac{1}{2}}^-,{\frac{3}{2}}^-,{\frac{5}{2}}^-$ for the vector-baryon (${\frac{3}{2}}^+$) channels. We look for poles of the states and evaluate the couplings to the different channels. The couplings obtained for the open channels can serve as a guide to see in which reaction the obtained states are more likely to be observed., Comment: 18 pages, 1 figures, accepted for publication in Phys.Rev.D

Tidal disruption events (TDEs) occur when a star passes too close to a supermassive black hole and is destroyed by tidal gravitational forces. Radio observations of TDEs trace synchrotron emission from outflowing material that may be ejected from the inner regions of the accretion flow around the SMBH or by the tidal debris stream. Radio detections of tidal disruption events are rare, but provide crucial information about the launching of jets and outflows from supermassive black holes and the circumnuclear environment in galaxies. Here we present the radio detection of the TDE AT2020opy, including three epochs of radio observations taken with the Karl G. Jansky's Very Large Array (VLA), MeerKAT, and upgraded Giant Metrewave Radio telescope. AT2020opy is the most distant thermal TDE with radio emission reported to date, and from modelling the evolving synchrotron spectra we deduce that the host galaxy has a more dense circumnuclear medium than other thermal TDEs detected in the radio band. Based on an equipartition analysis of the synchrotron spectral properties of the event, we conclude that the radio-emitting outflow was likely launched approximately at the time of, or just after, the initial optical flare. We find no evidence for relativistic motion of the outflow. The high luminosity of this event supports that a dense circumnuclear medium of the host galaxy produces brighter radio emission that rises to a peak more quickly than in galaxies with lower central densities., Comment: 8 pages, submitted to MNRAS

As a major approach to looking for life beyond the Earth, the search for extraterrestrial intelligence (SETI) is committed to detecting technosignatures such as engineered radio signals that are indicative of technologically capable life. In this paper, we report a targeted SETI campaign employing an observation strategy named multi-beam coincidence matching (MBCM) at the Five-hundred-meter Aperture Spherical radio Telescope (FAST) towards 33 known exoplanet systems, searching for ETI narrow-band drifting signals across 1.05-1.45 GHz in two orthogonal linear polarization directions separately. A signal at 1140.604 MHz detected from the observation towards Kepler-438 originally peaked our interest because its features are roughly consistent with assumed ETI technosignatures. However, evidences such as its polarization characteristics are almost able to eliminate the possibility of an extraterrestrial origin. Our observations achieve an unprecedented sensitivity since the minimum equivalent isotropic radiated power (EIRP) we are able to detect reaches 1.48 x10^9 W., Comment: 17 pages, 12 figures, matches the published version

Recent studies in lossy compression show that distortion and perceptual quality are at odds with each other, which put forward the tradeoff between distortion and perception (D-P). Intuitively, to attain different perceptual quality, different decoders have to be trained. In this paper, we present a nontrivial finding that only two decoders are sufficient for optimally achieving arbitrary (an infinite number of different) D-P tradeoff. We prove that arbitrary points of the D-P tradeoff bound can be achieved by a simple linear interpolation between the outputs of a minimum MSE decoder and a specifically constructed perfect perceptual decoder. Meanwhile, the perceptual quality (in terms of the squared Wasserstein-2 distance metric) can be quantitatively controlled by the interpolation factor. Furthermore, to construct a perfect perceptual decoder, we propose two theoretically optimal training frameworks. The new frameworks are different from the distortion-plus-adversarial loss based heuristic framework widely used in existing methods, which are not only theoretically optimal but also can yield state-of-the-art performance in practical perceptual decoding. Finally, we validate our theoretical finding and demonstrate the superiority of our frameworks via experiments. Code is available at: https://github.com/ZeyuYan/Controllable-Perceptual-Compression, Comment: ICML 2022

The contributions for the kaon pair from the intermediate states $\rho(1450)^+$ and $\rho(1700)^+$ in the decays $B^+ \to \bar{D}^0 K^+ \bar{K}^0$, $B^0 \to D^- K^+ \bar{K}^0$, and $B_s^0 \to D_s^-K^+ \bar{K}^0$ are analyzed within the perturbative QCD factorization approach. The decay amplitudes for all concerned decays in this work are dominated by the factorizable Feynman diagrams with the emission of the kaon pair, and the charged $\rho$ mesons should be of great importance in the $KK$ channel of the related three-body $B$ decays. Moreover, these quasi-two-body decays are CKM-favored, and the relevant branching ratios are predicted to be in the order of $10^{-5}$, which have the potential to be measured by experiments. It is also shown that the contributions of the subprocesses $\rho(1450, 1700)^+ \to KK$ for the three-body $B$ meson decays are considerable according to the total three-body branching fractions presented by Belle. Therefore, the decays $B^+ \to \bar{D}^0 K^+ \bar{K}^0$, $B^0 \to D^- K^+ \bar{K}^0$, and $B_s^0 \to D_s^-K^+ \bar{K}^0$ can be employed to study the properties of $\rho(1450)$ and $\rho(1700)$ in the LHCb and Belle-II experiments., Comment: 9 pages and 1 figure, matching the published version

Multi-label learning in the presence of missing labels (MLML) is a challenging problem. Existing methods mainly focus on the design of network structures or training schemes, which increase the complexity of implementation. This work seeks to fulfill the potential of loss function in MLML without increasing the procedure and complexity. Toward this end, we propose two simple yet effective methods via robust loss design based on an observation that a model can identify missing labels during training with a high precision. The first is a novel robust loss for negatives, namely the Hill loss, which re-weights negatives in the shape of a hill to alleviate the effect of false negatives. The second is a self-paced loss correction (SPLC) method, which uses a loss derived from the maximum likelihood criterion under an approximate distribution of missing labels. Comprehensive experiments on a vast range of multi-label image classification datasets demonstrate that our methods can remarkably boost the performance of MLML and achieve new state-of-the-art loss functions in MLML.

In scanning tunneling microscopy induced luminescence (STML), the photon counting is measured to reflect the single-molecule properties, e.g., the first molecular excited state. The energy of the first excited state is typically determined by a rising position of the photon counting as a function of the bias voltage between the tip and the substrate. It remains a challenge to determine the precise rise position of the current due to the possible experimental noise. In this work, we propose an alternating current version of STML to resolve the fine structures in the photon counting measurement. The measured photon counting and the current at the long-time limit show a sinusoidal oscillation. The zero-frequency component of the current shows knee points at the precise voltage as the fraction of the detuning between the molecular gap and the DC component of bias voltage. We propose to measure the energy level with discontinuity of the first derivative of such zero-frequency component. The current method will extend the application of STML in terms of measuring molecular properties.