Search

Imitation learning has achieved great success in many sequential decision-making tasks, in which a neural agent is learned by imitating collected human demonstrations. However, existing algorithms typically require a large number of high-quality demonstrations that are difficult and expensive to collect. Usually, a trade-off needs to be made between demonstration quality and quantity in practice. Targeting this problem, in this work we consider the imitation of sub-optimal demonstrations, with both a small clean demonstration set and a large noisy set. Some pioneering works have been proposed, but they suffer from many limitations, e.g., assuming a demonstration to be of the same optimality throughout time steps and failing to provide any interpretation w.r.t knowledge learned from the noisy set. Addressing these problems, we propose {\method} by evaluating and imitating at the sub-demonstration level, encoding action primitives of varying quality into different skills. Concretely, {\method} consists of a high-level controller to discover skills and a skill-conditioned module to capture action-taking policies, and is trained following a two-phase pipeline by first discovering skills with all demonstrations and then adapting the controller to only the clean set. A mutual-information-based regularization and a dynamic sub-demonstration optimality estimator are designed to promote disentanglement in the skill space. Extensive experiments are conducted over two gym environments and a real-world healthcare dataset to demonstrate the superiority of {\method} in learning from sub-optimal demonstrations and its improved interpretability by examining learned skills.

The endoplasmic reticulum (ER) is the main storage site of Zn2+, and Zn2+ plays an important role in regulating ER homeostasis. Therefore, we designed and synthesized a ratiometric fluorescent Zn2+ probe ER-Zn targeting ER stress. The probe displayed a specific Zn2+ induced blue shift at the spectral maximum values of excitation (80 nm) and emission (30 nm). The ratio imaging capability of Zn2+ under dual excitation mode can be applied not only to quantitative and reversible detection of exogenous Zn2+, but also the observation of the Zn2+ level change under ER stress, elucidating the different behaviors of Zn2+ release in ER stimulated by tunicamycin and thapsigargin. Additionally, the NIR imaging capability of ER-Zn provides an important basis for further research on animal models and is expected to realize the visualization and treatment of ER stress-related diseases through the regulation of ER stress by Zn2+. We envision that this probe can be applied to screen drugs for diseases related to ER stress regulation.

In this paper, we propose an ordered time series classification framework that is robust against missing classes in the training data, i.e., during testing we can prescribe classes that are missing during training. This framework relies on two main components: (1) our newly proposed ordinal-quadruplet loss, which forces the model to learn latent representation while preserving the ordinal relation among labels, (2) testing procedure, which utilizes the property of latent representation (order preservation). We conduct experiments based on real world multivariate time series data and show the significant improvement in the prediction of missing labels even with 40% of the classes are missing from training. Compared with the well-known triplet loss optimization augmented with interpolation for missing information, in some cases, we nearly double the accuracy.

Despite the fact that many anomaly detection approaches have been developed for multivariate time series data, limited effort has been made on federated settings in which multivariate time series data are heterogeneously distributed among different edge devices while data sharing is prohibited. In this paper, we investigate the problem of federated unsupervised anomaly detection and present a Federated Exemplar-based Deep Neural Network (Fed-ExDNN) to conduct anomaly detection for multivariate time series data on different edge devices. Specifically, we first design an Exemplar-based Deep Neural network (ExDNN) to learn local time series representations based on their compatibility with an exemplar module which consists of hidden parameters learned to capture varieties of normal patterns on each edge device. Next, a constrained clustering mechanism (FedCC) is employed on the centralized server to align and aggregate the parameters of different local exemplar modules to obtain a unified global exemplar module. Finally, the global exemplar module is deployed together with a shared feature encoder to each edge device and anomaly detection is conducted by examining the compatibility of testing data to the exemplar module. Fed-ExDNN captures local normal time series patterns with ExDNN and aggregates these patterns by FedCC, and thus can handle the heterogeneous data distributed over different edge devices simultaneously. Thoroughly empirical studies on six public datasets show that ExDNN and Fed-ExDNN can outperform state-of-the-art anomaly detection algorithms and federated learning techniques.

Detecting abnormal activities in real-world surveillance videos is an important yet challenging task as the prior knowledge about video anomalies is usually limited or unavailable. Despite that many approaches have been developed to resolve this problem, few of them can capture the normal spatio-temporal patterns effectively and efficiently. Moreover, existing works seldom explicitly consider the local consistency at frame level and global coherence of temporal dynamics in video sequences. To this end, we propose Convolutional Transformer based Dual Discriminator Generative Adversarial Networks (CT-D2GAN) to perform unsupervised video anomaly detection. Specifically, we first present a convolutional transformer to perform future frame prediction. It contains three key components, i.e., a convolutional encoder to capture the spatial information of the input video clips, a temporal self-attention module to encode the temporal dynamics, and a convolutional decoder to integrate spatio-temporal features and predict the future frame. Next, a dual discriminator based adversarial training procedure, which jointly considers an image discriminator that can maintain the local consistency at frame-level and a video discriminator that can enforce the global coherence of temporal dynamics, is employed to enhance the future frame prediction. Finally, the prediction error is used to identify abnormal video frames. Thoroughly empirical studies on three public video anomaly detection datasets, i.e., UCSD Ped2, CUHK Avenue, and Shanghai Tech Campus, demonstrate the effectiveness of the proposed adversarial spatio-temporal modeling framework., Accepted for publication in the 29th ACM International Conference on Multimedia (ACMMM '21)

We present two mitochondria-immobilized fluorescent probes ECPI-12 and IVPI-12 for long-term mitochondria visualization and tracking regardless of MMP changes., Visualizing and tracking mitochondrial dynamic changes is crucially important in the fields of physiology, pathology and pharmacology. Traditional electrostatic-attraction based mitochondrial probes fail to visualize and track the changes due to their leakage from mitochondria when mitochondrial membrane potential (MMP) decreases. Reaction-based MitoTracker probes can realize visualization and tracking of mitochondria changes independent of MMP changes. However, such probes impair mitochondrial proteins and exhibit high cytotoxicity. Therefore, it still remains challenging to explore reaction-free and highly biocompatible probes for visualizing and tracking mitochondrial dynamics independent of MMP fluctuations. Herein we synthesized two reaction-free fluorescent mitochondrial probes ECPI-12 and IVPI-12 bearing a long C12-alkyl chain. These cationic probes can firmly immobilize in the mitochondrial inner membrane by strong hydrophobic interaction between the C12-alkyl chain and lipid bilayer, resulting in high specificity and long-term mitochondrial staining regardless of MMP changes. They also exhibit large two-photon absorption cross-sections and show deep penetration in live tissues in two-photon microscopy. Furthermore, they display excellent biocompatibility and realize in situ and real-time mitophagy tracking in live cells. These excellent properties could make ECPI-12 and IVPI-12 the first selective tools for long-term visualization and tracking of mitochondrial dynamics.

Question routing (QR) aims at recommending newly posted questions to the potential answerers who are most likely to answer the questions. The existing approaches that learn users' expertise from their past question-answering activities usually suffer from challenges in two aspects: 1) multi-faceted expertise and 2) temporal dynamics in the answering behavior. This paper proposes a novel temporal context-aware model in multiple granularities of temporal dynamics that concurrently address the above challenges. Specifically, the temporal context-aware attention characterizes the answerer's multi-faceted expertise in terms of the questions' semantic and temporal information simultaneously. Moreover, the design of the multi-shift and multi-resolution module enables our model to handle temporal impact on different time granularities. Extensive experiments on six datasets from different domains demonstrate that the proposed model significantly outperforms competitive baseline models.

This project integrates infrared and RGB imagery to produce dense 3D environment models reconstructed from multiple views. The resulting 3D map contains both thermal and RGB information which can be used in robotic fire-fighting applications to identify victims and active fire areas., Comment: Technical report submitted to 2013 DRS Student Infrared Imaging Competition

Brightfield and fluorescent imaging of whole brain sections are funda- mental tools of research in mouse brain study. As sectioning and imaging become more efficient, there is an increasing need to automate the post-processing of sec- tions for alignment and three dimensional visualization. There is a further need to facilitate the development of a digital atlas, i.e. a brain-wide map annotated with cell type and tract tracing data, which would allow the automatic registra- tion of images stacks to a common coordinate system. Currently, registration of slices requires manual identification of landmarks. In this work we describe the first steps in developing a semi-automated system to construct a histology at- las of mouse brainstem that combines atlas-guided annotation, landmark-based registration and atlas generation in an iterative framework. We describe an unsu- pervised approach for identifying and matching region and boundary landmarks, based on modelling texture. Experiments show that the detected landmarks corre- spond well with brain structures, and matching is robust under distortion. These results will serve as the basis for registration and atlas building., Comment: Submitted to MICCAI 2015

An anatomical atlas constitutes a quantitative description of the structural organization of an organism. Nissl-stained serial brain sections reveal cellular texture (cytoarchitecture) and are the gold standard for defining structures. As high-throughput techniques advance and automated registration becomes commonplace, the role of an atlas grows beyond a static map towards a repository where a wide variety of experiment data acquired from different subjects can be integrated in a common template defined in a standard coordinate system. For histological data (0.5 micron resolution), however, the reliance of intensity-based registration algorithms on downsampled pixel/voxel intensities (> 10 micron) and the resulting negligence of fine-scale textures means they often fail to accurately align structures whose boundaries are defined by cytoarchitecture rather than graylevel, such as many brainstem nuclei. The lack of a registration tool that strongly utilizes texture and a nucleus-level atlas to assist such registration has stifled comparison of results across experiments.We demonstrate a data-driven active atlas system that automatically aligns brains based on cytoarchitectural landmarks. Our approach combines discriminative texture detectors based on CNN features with a reference atlas that describes structure shapes as probabilistic volumes and their locations as Gaussian distributions. Histological serial sections are reconstructed in 3-D and converted to structure probability maps by classifiers trained to differentiate the texture inside versus outside each structure. Registration is achieved by maximization of the correlation between the probability maps and the reference atlas using a global affine transform followed by deformation interpolated from structure-specific rigid transforms. Initialized from annotation by expert neuroanatomists, the atlas is continuously refined after incorporation of new brains in a semi-supervised fashion.Based on automated registration of twelve specimens, we developed an atlas for adult murine brainstem which defines 28 distinct structures in both hemispheres. The system's utility in advancing brain circuitry study is demonstrated by the precise mapping of neuronal projections in cytoarchitecturally ill-defined regions across brains from different animals. Quantitative results showed that our approach produced accurate and confident registration and significantly reduces human labor.

While modern imaging technologies such as fMRI have opened exciting new possibilities for studying the brain in vivo, histological sections remain the best way to study the anatomy of the brain at the level of single neurons. The histological atlas changed little since 1909 and localizing brain regions is a still a labor intensive process performed only by experienced neuro-anatomists. Existing digital atlases such as the Allen Brain atlas are limited to low resolution images which cannot identify the detailed structure of the neurons. We have developed a digital atlas methodology that combines information about the 3D organization of the brain and the detailed texture of neurons in different structures. Using the methodology we developed an atlas for the mouse brainstem and mid-brain, two regions for which there are currently no good atlases. Our atlas is "active" in that it can be used to automatically align a histological stack to the atlas, thus reducing the work of the neuroanatomist., Comment: 8 pages, 10 figures, appeared in proceeding of MICCAI 2017

The MiR-146a/TRAF6/NF-κB axis is important for the regulation of hematopoiesis and the immune system. To identify the key axis that regulates benzene-induced hematotoxicity or even leukemia, we investigated miR-146a expression in human CD34

Nowadays, multivariate time series data are increasingly collected in various real world systems, e.g., power plants, wearable devices, etc. Anomaly detection and diagnosis in multivariate time series refer to identifying abnormal status in certain time steps and pinpointing the root causes. Building such a system, however, is challenging since it not only requires to capture the temporal dependency in each time series, but also need encode the inter-correlations between different pairs of time series. In addition, the system should be robust to noise and provide operators with different levels of anomaly scores based upon the severity of different incidents. Despite the fact that a number of unsupervised anomaly detection algorithms have been developed, few of them can jointly address these challenges. In this paper, we propose a Multi-Scale Convolutional Recurrent Encoder-Decoder (MSCRED), to perform anomaly detection and diagnosis in multivariate time series data. Specifically, MSCRED first constructs multi-scale (resolution) signature matrices to characterize multiple levels of the system statuses in different time steps. Subsequently, given the signature matrices, a convolutional encoder is employed to encode the inter-sensor (time series) correlations and an attention based Convolutional Long-Short Term Memory (ConvLSTM) network is developed to capture the temporal patterns. Finally, based upon the feature maps which encode the inter-sensor correlations and temporal information, a convolutional decoder is used to reconstruct the input signature matrices and the residual signature matrices are further utilized to detect and diagnose anomalies. Extensive empirical studies based on a synthetic dataset and a real power plant dataset demonstrate that MSCRED can outperform state-ofthe-art baseline methods.