Search

Irregular distribution in latent space causes posterior collapse, misalignment between posterior and prior, and ill-sampling problem in Variational Autoencoders (VAEs). In this paper, we introduce a novel adaptable three-stage Uniform Transformation (UT) module -- Gaussian Kernel Density Estimation (G-KDE) clustering, non-parametric Gaussian Mixture (GM) Modeling, and Probability Integral Transform (PIT) -- to address irregular latent distributions. By reconfiguring irregular distributions into a uniform distribution in the latent space, our approach significantly enhances the disentanglement and interpretability of latent representations, overcoming the limitation of traditional VAE models in capturing complex data structures. Empirical evaluations demonstrated the efficacy of our proposed UT module in improving disentanglement metrics across benchmark datasets -- dSprites and MNIST. Our findings suggest a promising direction for advancing representation learning techniques, with implication for future research in extending this framework to more sophisticated datasets and downstream tasks., Comment: Accepted by 2024 IEEE 20th International Conference on Automation Science and Engineering

Layered architectures have been widely used in robot systems. The majority of them implement planning and execution functions in separate layers. However, there still lacks a straightforward way to transit high-level tasks in the planning layer to the low-level motor commands in the execution layer. In order to tackle this challenge, we propose a novel approach to ground the manipulator primitive tasks to robot low-level actions using large language models (LLMs). We designed a program-function-like prompt based on the task frame formalism. In this way, we enable LLMs to generate position/force set-points for hybrid control. Evaluations over several state-of-the-art LLMs are provided., Comment: AAAI Fall Symposium on Unifying Representations for Robot Application Development, Arlington, VA, 2023

Symbolic regression searches for analytic expressions that accurately describe studied phenomena. The main attraction of this approach is that it returns an interpretable model that can be insightful to users. Historically, the majority of algorithms for symbolic regression have been based on evolutionary algorithms. However, there has been a recent surge of new proposals that instead utilize approaches such as enumeration algorithms, mixed linear integer programming, neural networks, and Bayesian optimization. In order to assess how well these new approaches behave on a set of common challenges often faced in real-world data, we hosted a competition at the 2022 Genetic and Evolutionary Computation Conference consisting of different synthetic and real-world datasets which were blind to entrants. For the real-world track, we assessed interpretability in a realistic way by using a domain expert to judge the trustworthiness of candidate models.We present an in-depth analysis of the results obtained in this competition, discuss current challenges of symbolic regression algorithms and highlight possible improvements for future competitions., Comment: 13 pages, 13 figures, submitted to IEEE Transactions on Evolutionary Computation

Nowadays, the behavior tree is gaining popularity as a representation for robot tasks due to its modularity and reusability. Designing behavior-tree tasks manually is time-consuming for robot end-users, thus there is a need for investigating automatic behavior-tree-based task generation. Prior behavior-tree-based task generation approaches focus on fixed primitive tasks and lack generalizability to new task domains. To cope with this issue, we propose a novel behavior-tree-based task generation approach that utilizes state-of-the-art large language models. We propose a Phase-Step prompt design that enables a hierarchical-structured robot task generation and further integrate it with behavior-tree-embedding-based search to set up the appropriate prompt. In this way, we enable an automatic and cross-domain behavior-tree task generation. Our behavior-tree-based task generation approach does not require a set of pre-defined primitive tasks. End-users only need to describe an abstract desired task and our proposed approach can swiftly generate the corresponding behavior tree. A full-process case study is provided to demonstrate our proposed approach. An ablation study is conducted to evaluate the effectiveness of our Phase-Step prompts. Assessment on Phase-Step prompts and the limitation of large language models are presented and discussed., Comment: The extended abstract of this paper is accepted in AAAI 2023 Spring Symposium on Challenges Requiring the Combination of Machine Learning and Knowledge Engineering (AAAI-MAKE 2023)

In this paper, we address the Online Unsupervised Domain Adaptation (OUDA) problem and propose a novel multi-stage framework to solve real-world situations when the target data are unlabeled and arriving online sequentially in batches. To project the data from the source and the target domains to a common subspace and manipulate the projected data in real-time, our proposed framework institutes a novel method, called an Incremental Computation of Mean-Subspace (ICMS) technique, which computes an approximation of mean-target subspace on a Grassmann manifold and is proven to be a close approximate to the Karcher mean. Furthermore, the transformation matrix computed from the mean-target subspace is applied to the next target data in the recursive-feedback stage, aligning the target data closer to the source domain. The computation of transformation matrix and the prediction of next-target subspace leverage the performance of the recursive-feedback stage by considering the cumulative temporal dependency among the flow of the target subspace on the Grassmann manifold. The labels of the transformed target data are predicted by the pre-trained source classifier, then the classifier is updated by the transformed data and predicted labels. Extensive experiments on six datasets were conducted to investigate in depth the effect and contribution of each stage in our proposed framework and its performance over previous approaches in terms of classification accuracy and computational speed. In addition, the experiments on traditional manifold-based learning models and neural-network-based learning models demonstrated the applicability of our proposed framework for various types of learning models.

A safety-critical measure of legged locomotion performance is a robot's ability to track its desired time-varying position trajectory in an environment, which is herein termed as "global-position tracking". This paper introduces a nonlinear control approach that achieves asymptotic global-position tracking for three-dimensional (3-D) bipedal robot walking. Designing a global-position tracking controller presents a challenging problem due to the complex hybrid robot model and the time-varying desired global-position trajectory. Towards tackling this problem, the first main contribution is the construction of impact invariance to ensure all desired trajectories respect the foot-landing impact dynamics, which is a necessary condition for realizing asymptotic tracking of hybrid walking systems. Thanks to their independence of the desired global position, these conditions can be exploited to decouple the higher-level planning of the global position and the lower-level planning of the remaining trajectories, thereby greatly alleviating the computational burden of motion planning. The second main contribution is the Lyapunov-based stability analysis of the hybrid closed-loop system, which produces sufficient conditions to guide the controller design for achieving asymptotic global-position tracking during fully actuated walking. Simulations and experiments on a 3-D bipedal robot with twenty revolute joints confirm the validity of the proposed control approach in guaranteeing accurate tracking., Comment: 16 pages

In this paper, we extend the traditional few-shot learning (FSL) problem to the situation when the source-domain data is not accessible but only high-level information in the form of class prototypes is available. This limited information setup for the FSL problem deserves much attention due to its implication of privacy-preserving inaccessibility to the source-domain data but it has rarely been addressed before. Because of limited training data, we propose a non-parametric approach to this FSL problem by assuming that all the class prototypes are structurally arranged on a manifold. Accordingly, we estimate the novel-class prototype locations by projecting the few-shot samples onto the average of the subspaces on which the surrounding classes lie. During classification, we again exploit the structural arrangement of the categories by inducing a Markov chain on the graph constructed with the class prototypes. This manifold distance obtained using the Markov chain is expected to produce better results compared to a traditional nearest-neighbor-based Euclidean distance. To evaluate our proposed framework, we have tested it on two image datasets - the large-scale ImageNet and the small-scale but fine-grained CUB-200. We have also studied parameter sensitivity to better understand our framework., Comment: International Symposium on Visual Computing (ISVC), 2020

The parity-transfer $({}^{16}{\rm O},{}^{16}{\rm F}(0^-,{\rm g.s.}))$ reaction is presented as a new probe for investigating isovector $0^-$ states in nuclei. The properties of $0^-$ states provide a stringent test of the threshold density for pion condensation in nuclear matter. Utilizing a $0^+ \rightarrow 0^-$ transition in the projectile, the parity-transfer reaction transfers an internal parity to a target nucleus, resulting in a unique sensitivity to unnatural-parity states. Consequently, the selectivity for $0^-$ states is higher than in other reactions employed to date. The probe was applied to a study of the $0^-$ states in ${}^{12}{\rm B}$ via the ${}^{12}{\rm C}({}^{16}{\rm O},{}^{16}{\rm F}(0^-,{\rm g.s.}))$ reaction at $247~{\rm MeV/u}$. The excitation energy spectra were deduced by detecting the ${}^{15}{\rm O}+p$ pair produced in the decay of the ${}^{16}{\rm F}$ ejectile. A known $0^-$ state at $E_x = 9.3~{\rm MeV}$ was observed with an unprecedentedly high signal-to-noise ratio. The data also revealed new candidates of $0^-$ states at $E_x=6.6 \pm 0.4$ and $14.8 \pm 0.3~{\rm MeV}$. The results demonstrate the high efficiency of $0^-$ state detection by the parity-transfer reaction., Comment: Submitted to Phys. Rev. Lett

In this paper, we address the Online Unsupervised Domain Adaptation (OUDA) problem, where the target data are unlabelled and arriving sequentially. The traditional methods on the OUDA problem mainly focus on transforming each arriving target data to the source domain, and they do not sufficiently consider the temporal coherency and accumulative statistics among the arriving target data. We propose a multi-step framework for the OUDA problem, which institutes a novel method to compute the mean-target subspace inspired by the geometrical interpretation on the Euclidean space. This mean-target subspace contains accumulative temporal information among the arrived target data. Moreover, the transformation matrix computed from the mean-target subspace is applied to the next target data as a preprocessing step, aligning the target data closer to the source domain. Experiments on four datasets demonstrated the contribution of each step in our proposed multi-step OUDA framework and its performance over previous approaches., Comment: To appear in ICASSP 2020. Copyright 2020 IEEE

This paper proposes a multi-layer neural network structure for few-shot image recognition of novel categories. The proposed multi-layer neural network architecture encodes transferable knowledge extracted from a large annotated dataset of base categories. This architecture is then applied to novel categories containing only a few samples. The transfer of knowledge is carried out at the feature-extraction and the classification levels distributed across the two training stages. In the first-training stage, we introduce the relative feature to capture the structure of the data as well as obtain a low-dimensional discriminative space. Secondly, we account for the variable variance of different categories by using a network to predict the variance of each class. Classification is then performed by computing the Mahalanobis distance to the mean-class representation in contrast to previous approaches that used the Euclidean distance. In the second-training stage, a category-agnostic mapping is learned from the mean-sample representation to its corresponding class-prototype representation. This is because the mean-sample representation may not accurately represent the novel category prototype. Finally, we evaluate the proposed network structure on four standard few-shot image recognition datasets, where our proposed few-shot learning system produces competitive performance compared to previous work. We also extensively studied and analyzed the contribution of each component of our proposed framework., Comment: To Appear in IEEE Transactions on Image Processing

Zero-shot learning (ZSL) for image classification focuses on recognizing novel categories that have no labeled data available for training. The learning is generally carried out with the help of mid-level semantic descriptors associated with each class. This semantic-descriptor space is generally shared by both seen and unseen categories. However, ZSL suffers from hubness, domain discrepancy and biased-ness towards seen classes. To tackle these problems, we propose a three-step approach to zero-shot learning. Firstly, a mapping is learned from the semantic-descriptor space to the image-feature space. This mapping learns to minimize both one-to-one and pairwise distances between semantic embeddings and the image features of the corresponding classes. Secondly, we propose test-time domain adaptation to adapt the semantic embedding of the unseen classes to the test data. This is achieved by finding correspondences between the semantic descriptors and the image features. Thirdly, we propose scaled calibration on the classification scores of the seen classes. This is necessary because the ZSL model is biased towards seen classes as the unseen classes are not used in the training. Finally, to validate the proposed three-step approach, we performed experiments on four benchmark datasets where the proposed method outperformed previous results. We also studied and analyzed the performance of each component of our proposed ZSL framework., Comment: International Joint Conference on Neural Networks (IJCNN) 2019. Copyright 2019 IEEE

The neutron-shell structure of $^{25}$F was studied using quasi-free (p,2p) knockout reaction at 270A MeV in inverse kinematics. The sum of spectroscopic factors of $\pi$0d$_{5/2}$ orbital is found to be $1.0 \pm 0.3$. However, the spectroscopic factor for the ground-state to ground-state transition ($^{25}$F, $^{24}$O$_{g.s.}$) is only $0.36\pm 0.13$, and $^{24}$O excited states are produced from the 0d$_{5/2}$ proton knockout. The result shows that the $^{24}$O core of $^{25}$F nucleus significantly differs from a free $^{24}$O nucleus, and the core consists of 35% $^{24}$O$_{g.s}$. and 65% excited $^{24}$O., Comment: 5 pages, 5 figures

Domain adaptation (DA) addresses the real-world image classification problem of discrepancy between training (source) and testing (target) data distributions. We propose an unsupervised DA method that considers the presence of only unlabelled data in the target domain. Our approach centers on finding matches between samples of the source and target domains. The matches are obtained by treating the source and target domains as hyper-graphs and carrying out a class-regularized hyper-graph matching using first-, second- and third-order similarities between the graphs. We have also developed a computationally efficient algorithm by initially selecting a subset of the samples to construct a graph and then developing a customized optimization routine for graph-matching based on Conditional Gradient and Alternating Direction Multiplier Method. This allows the proposed method to be used widely. We also performed a set of experiments on standard object recognition datasets to validate the effectiveness of our framework over state-of-the-art approaches., Comment: Final version appeared in IEEE International Conference on Image Processing 2018

The assumption that training and testing samples are generated from the same distribution does not always hold for real-world machine-learning applications. The procedure of tackling this discrepancy between the training (source) and testing (target) domains is known as domain adaptation. We propose an unsupervised version of domain adaptation that considers the presence of only unlabelled data in the target domain. Our approach centers on finding correspondences between samples of each domain. The correspondences are obtained by treating the source and target samples as graphs and using a convex criterion to match them. The criteria used are first-order and second-order similarities between the graphs as well as a class-based regularization. We have also developed a computationally efficient routine for the convex optimization, thus allowing the proposed method to be used widely. To verify the effectiveness of the proposed method, computer simulations were conducted on synthetic, image classification and sentiment classification datasets. Results validated that the proposed local sample-to-sample matching method out-performs traditional moment-matching methods and is competitive with respect to current local domain-adaptation methods., Comment: Final version appeared in Engineering Applications of Artificial Intelligence. Mostly, the related work in this version is different from the published version

It is a well-known fact that a cluster of nucleons can be formed in the interior of an atomic nucleus, and such clusters may occupy molecular-like orbitals, showing characteristics similar to normal molecules consisting of atoms. Chemical molecules having a linear alignment are commonly seen in nature, such as carbon dioxide. A similar linear alignment of the nuclear clusters, referred to as linear-chain cluster state (LCCS), has been studied since the 1950s, however, up to now there is no clear experimental evidence demonstrating the existence of such a state. Recently, it was proposed that an excess of neutrons may offer just such a stabilizing mechanism, revitalizing interest in the nuclear LCCS, specifically with predictions for their emergence in neutron-rich carbon isotopes. Here we present the experimental observation of {\alpha}-cluster states in the radioactive 14C nucleus. Using the 10Be+{\alpha} resonant scattering method with a radioactive beam, we observed a series of levels which completely agree with theoretically predicted levels having an explicit linear-chain cluster configuration. We regard this as the first strong indication of the linear-chain clustered nucleus., Comment: 15 pages, 3 figures

We study for the first time the internal structure of 140Te through the beta-delayed gamma-ray spectroscopy of 140Sb. The very neutron-rich 140Sb, Z = 51 and N = 89, ions were produced by the in-flight fission of 238U beam on a 9Be target at 345 MeV per nucleon at the Radioactive Ion Beam Factory, RIKEN. The half-life and spin-parity of 140Sb are reported as 124(30) ms and (4-), respectively. In addition to the excited states of 140Te produced by the beta-decay branch, the beta-delayed one-neutron and two-neutron emission branches were also established. By identifying the first 2+ and 4+ excited states of 140Te, we found that Te isotopes persist their vibrator character with E(4+)/E(2+) = 2. We discuss the distinctive features manifest in this region, such as valence neutron symmetry and asymmetry, revealed in pairs of isotopes with the same neutron holes and particles with respect to N = 82., Comment: 9 pages, 6 figures, EURICA campaign

The $^{14}$O($\alpha$,$p$)$^{17}$F reaction is one of the key reactions involved in the breakout from the hot-CNO cycle to the rp-process in type I x-ray bursts (XRBs). The resonant properties in the compound nucleus $^{18}$Ne have been investigated through resonant elastic scattering of $^{17}$F+$p$. The radioactive $^{17}$F beam was separated by the CNS Radioactive Ion Beam separator (CRIB) and bombarded a thick H$_2$ gas target at 3.6 MeV/nucleon. The recoiling light particles were measured by three ${\Delta}$E-E silicon telescopes at laboratory angles of $\theta$$_{lab}$$\approx$3$^\circ$, 10$^\circ$ and 18$^\circ$, respectively. Five resonances at $E_{x}$=6.15, 6.28, 6.35, 6.85, and 7.05 MeV were observed in the excitation functions, and their spin-parities have been determined based on an $R$-matrix analysis. In particular, $J^{\pi}$=1$^-$ was firmly assigned to the 6.15-MeV state which dominates the thermonuclear $^{14}$O($\alpha$,$p$)$^{17}$F rate below 2 GK. As well, a possible new excited state in $^{18}$Ne was observed at $E_{x}$=6.85$\pm$0.11 MeV with tentative $J$=0 assignment. This state could be the analog state of the 6.880 MeV (0$^{-}$) level in the mirror nucleus $^{18}$O, or a bandhead state (0$^+$) of the six-particle four-hole (6$p$-4$h$) band. A new thermonuclear $^{14}$O($\alpha$,$p$)$^{17}$F rate has been determined, and the astrophysical impact of multiple recent rates has been examined using an XRB model. Contrary to previous expectations, we find only modest impact on predicted nuclear energy generation rates from using reaction rates differing by up to several orders of magnitude., Comment: 8 pages, 5 figures

The $^{14}$O($\alpha$,$p$)$^{17}$F reaction is one of the key reactions involved in the breakout from the hot-CNO cycle to the rp-process in type I x-ray bursts. The resonant properties in the compound nucleus $^{18}$Ne have been investigated through resonant elastic scattering of $^{17}$F+$p$. The radioactive $^{17}$F beam was separated by the CNS Radioactive Ion Beam separator (CRIB) and bombarded a thick H$_2$ gas target at 3.6 MeV/nucleon. The recoiling light particles were measured by using three ${\Delta}$E-E silicon telescopes at laboratory angles of $\theta$$_{lab}$$\approx$3$^\circ$, 10$^\circ$ and 18$^\circ$, respectively. Five resonances at $E_{x}$=6.15, 6.28, 6.35, 6.85, and 7.05 MeV were observed in the excitation functions. Based on an $R$-matrix analysis, $J^{\pi}$=1$^-$ was firmly assigned to the 6.15-MeV state. This state dominates the thermonuclear $^{14}$O($\alpha$,$p$)$^{17}$F rate below 1 GK. We have also confirmed the existence and spin-parities of three states between 6.1 and 6.4 MeV. As well, a possible new excited state in $^{18}$Ne was observed at $E_{x}$=6.85$\pm$0.11 MeV and tentatively assigned as $J$=0. This state could be the analog state of the 6.880 MeV (0$^{-}$) level in the mirror nucleus $^{18}$O, or a bandhead state (0$^+$) of the six-particle four-hole (6$p$-4$h$) band. A new thermonuclear rate of the $^{14}$O($\alpha$,$p$)$^{17}$F reaction has been determined, and its astrophysical impact has been examined within the framework of one-zone x-ray burst postprocessing calculations.

Growth on transition metal substrates is becoming a method of choice to prepare large-area graphene foils. In the case of nickel, where carbon has a significant solubility, such a growth process includes at least two elementary steps: (1) carbon dissolution into the metal, and (2) graphene precipitation at the surface. Here, we dissolve calibrated amounts of carbon in nickel films, using carbon ion implantation, and annealing at 725 \circ or 900 \circ. We then use transmission electron microscopy to analyse the precipitation process in detail: the latter appears to imply carbon diffusion over large distances and at least two distinct microscopic mechanisms.

Exposing single-walled carbon nanotubes to room temperature UV-generated ozone leads to an irreversible increase in their electrical resistance. We demonstrate that the increased resistance is due to ozone oxidation on the sidewalls of the nanotubes rather than at the end caps. Raman and x-ray photoelectron spectroscopy show an increase in the defect density due to the oxidation of the nanotubes. Using ultraviolet photoelectron spectroscopy we show that these defects represent the removal of pi-conjugated electron states near the Fermi level, leading to the observed increase in electrical resistance. Oxidation of carbon nanotubes is an important first step in many chemical functionalization processes. Since the oxidation rate is controllable with short exposures, UV-generated ozone offers the potential for use as a low-thermal budget processing tool., Comment: 28 pages, 9 figures

Three-dimensional (3D) micro-electromagnets were developed to control particle motion in magnetic field landscapes in vacuum near a chip. Multiple layers of micron-scale conductors separated by transparent insulators create particle containers with deep, symmetric and time-dependent potentials, suitable for integration in quantum circuits. Single and coupled multiple 3D traps, integrated 3D traps with a mirror, 3D guides for neutral particles with spin, and traps for electrons in vacuum are described., Comment: 15 pages, 1 table, 3 figures, Accepted for publication in Phys. Rev. B

Cryogenically cooled microelectromagnet mirrors were used to reflect a cloud of free-falling laser-cooled 85Rb atoms at normal incidence. The mirrors consisted of microfabricated current-carrying Au wires in a periodic serpentine pattern on a sapphire substrate. The fluorescence from the atomic cloud was imaged after it had bounced off a mirror. The transverse width of the cloud reached a local minimum at an optimal current corresponding to minimum mirror roughness. A distinct increase in roughness was found for mirror configurations with even versus odd number of lines. These observations confirm theoretical predictions., Comment: Physical Review A, in print; 11 pages, 4 figures

We demonstrate the guiding of neutral atoms by the magnetic fields due to microfabricated current-carrying wires on a chip. Atoms are guided along a magnetic field minimum parallel to and above the current-carrying wires. Two waveguide configurations are demonstrated: one using two wires with an external magnetic field, and a second using four wires without an external field. These waveguide geometries can be extended to integrated atom optics circuits, including beamsplitters., Comment: 11 pages, 4 figures

Recent experimental studies focusing on the morphological properties of surfaces eroded by ion-bombardment report the observation of self-affine fractal surfaces, while others provide evidence about the development of a periodic ripple structure. To explain these discrepancies we derive a stochastic growth equation that describes the evolution of surfaces eroded by ion bombardment. The coefficients appearing in the equation can be calculated explicitly in terms of the physical parameters characterizing the sputtering process. Exploring the connection between the ion-sputtering problem and the Kardar-Parisi-Zhang and Kuramoto-Sivashinsky equations, we find that morphological transitions may take place when experimental parameters, such as the angle of incidence of the incoming ions or their average penetration depth, are varied. Furthermore, the discussed methods allow us to calculate analytically the ion-induced surface diffusion coefficient, that can be compared with experiments. Finally, we use numerical simulations of a one dimensional sputtering model to investigate certain aspects of the ripple formation and roughening., Comment: 20 pages, LaTeX, 5 ps figures, contribution to the 4th CTP Workshop on Statistical Physics "Dynamics of Fluctuating Interfaces and Related Phenomena", Seoul National University, Seoul, Korea, January 27-31, 1997

The $^{174}$Yb($^{29}$Si,5n) reaction at 148 MeV with thin targets was used to populate high-angular momentum states in $^{198}$Po. Resulting $\gamma$ rays were observed with Gammasphere. A weakly-populated superdeformed band of 10 $\gamma$-ray transitions was found and has been assigned to $^{198}$Po. This is the first observation of a SD band in the $A \approx 190$ region in a nucleus with $Z > 83$. The ${\cal J}^{(2)}$ of the new band is very similar to those of the yrast SD bands in $^{194}$Hg and $^{196}$Pb. The intensity profile suggests that this band is populated through states close to where the SD band crosses the yrast line and the angular momentum at which the fission process dominates., Comment: 10 pages, revtex, 2 figs. available on request, submitted to Phys. Rev. C. (Rapid Communications)