Offline reinforcement learning (RL) faces a significant challenge of distribution shift. Model-free offline RL penalizes the Q value for out-of-distribution (OOD) data or constrains the policy closed to the behavior policy to tackle this problem, but this inhibits the exploration of the OOD region. Model-based offline RL, which uses the trained environment model to generate more OOD data and performs conservative policy optimization within that model, has become an effective method for this problem. However, the current model-based algorithms rarely consider agent robustness when incorporating conservatism into policy. Therefore, the new model-based offline algorithm with a conservative Bellman operator (MICRO) is proposed. This method trades off performance and robustness via introducing the robust Bellman operator into the algorithm. Compared with previous model-based algorithms with robust adversarial models, MICRO can significantly reduce the computation cost by only choosing the minimal Q value in the state uncertainty set. Extensive experiments demonstrate that MICRO outperforms prior RL algorithms in offline RL benchmark and is considerably robust to adversarial perturbations.
相關內容
Learning control from pixels is difficult for reinforcement learning (RL) agents because representation learning and policy learning are intertwined. Previous approaches remedy this issue with auxiliary representation learning tasks, but they either do not consider the temporal aspect of the problem or only consider single-step transitions, which may cause learning inefficiencies if important environmental changes take many steps to manifest. We propose Hierarchical $k$-Step Latent (HKSL), an auxiliary task that learns multiple representations via a hierarchy of forward models that learn to communicate and an ensemble of $n$-step critics that all operate at varying magnitudes of step skipping. We evaluate HKSL in a suite of 30 robotic control tasks with and without distractors and a task of our creation. We find that HKSL either converges to higher or optimal episodic returns more quickly than several alternative representation learning approaches. Furthermore, we find that HKSL's representations capture task-relevant details accurately across timescales (even in the presence of distractors) and that communication channels between hierarchy levels organize information based on both sides of the communication process, both of which improve sample efficiency.
Deep learning methods can not only detect false data injection attacks (FDIA) but also locate attacks of FDIA. Although adversarial false data injection attacks (AFDIA) based on deep learning vulnerabilities have been studied in the field of single-label FDIA detection, the adversarial attack and defense against multi-label FDIA locational detection are still not involved. To bridge this gap, this paper first explores the multi-label adversarial example attacks against multi-label FDIA locational detectors and proposes a general multi-label adversarial attack framework, namely muLti-labEl adverSarial falSe data injectiON attack (LESSON). The proposed LESSON attack framework includes three key designs, namely Perturbing State Variables, Tailored Loss Function Design, and Change of Variables, which can help find suitable multi-label adversarial perturbations within the physical constraints to circumvent both Bad Data Detection (BDD) and Neural Attack Location (NAL). Four typical LESSON attacks based on the proposed framework and two dimensions of attack objectives are examined, and the experimental results demonstrate the effectiveness of the proposed attack framework, posing serious and pressing security concerns in smart grids.
Unsupervised learning of object-centric representations in dynamic visual scenes is challenging. Unlike most previous approaches that learn to decompose 2D images, we present DynaVol, a 3D scene generative model that unifies geometric structures and object-centric learning in a differentiable volume rendering framework. The key idea is to perform object-centric voxelization to capture the 3D nature of the scene, which infers the probability distribution over objects at individual spatial locations. These voxel features evolve over time through a canonical-space deformation function, forming the basis for global representation learning via slot attention. The voxel features and global features are complementary and are both leveraged by a compositional NeRF decoder for volume rendering. DynaVol remarkably outperforms existing approaches for unsupervised dynamic scene decomposition. Once trained, the explicitly meaningful voxel features enable additional capabilities that 2D scene decomposition methods cannot achieve: it is possible to freely edit the geometric shapes or manipulate the motion trajectories of the objects.
A domain shift occurs when training (source) and test (target) data diverge in their distribution. Test-time adaptation (TTA) addresses the domain shift problem, aiming to adopt a trained model on the source domain to the target domain in a scenario where only a well-trained source model and unlabeled target data are available. In this scenario, handling false labels in the target domain is crucial because they negatively impact the model performance. To deal with this problem, we propose to utilize cluster structure (i.e., {`Clean'} and {`Noisy'} regions within each cluster) in the target domain formulated by the source model. Given an initial clustering of target samples, we first partition clusters into {`Clean'} and {`Noisy'} regions defined based on cluster prototype (i.e., centroid of each cluster). As these regions have totally different distributions of the true pseudo-labels, we adopt distinct training strategies for the clean and noisy regions: we selectively train the target with clean pseudo-labels in the clean region, whereas we introduce mixup inputs representing intermediate features between clean and noisy regions to increase the compactness of the cluster. We conducted extensive experiments on multiple datasets in online/offline TTA settings, whose results demonstrate that our method, {CNA-TTA}, achieves state-of-the-art for most cases.
Multi-task reinforcement learning endeavors to accomplish a set of different tasks with a single policy. To enhance data efficiency by sharing parameters across multiple tasks, a common practice segments the network into distinct modules and trains a routing network to recombine these modules into task-specific policies. However, existing routing approaches employ a fixed number of modules for all tasks, neglecting that tasks with varying difficulties commonly require varying amounts of knowledge. This work presents a Dynamic Depth Routing (D2R) framework, which learns strategic skipping of certain intermediate modules, thereby flexibly choosing different numbers of modules for each task. Under this framework, we further introduce a ResRouting method to address the issue of disparate routing paths between behavior and target policies during off-policy training. In addition, we design an automatic route-balancing mechanism to encourage continued routing exploration for unmastered tasks without disturbing the routing of mastered ones. We conduct extensive experiments on various robotics manipulation tasks in the Meta-World benchmark, where D2R achieves state-of-the-art performance with significantly improved learning efficiency.
We propose Compact and Swift Segmenting 3D Gaussians(CoSSegGaussians), a method for compact 3D-consistent scene segmentation at fast rendering speed with only RGB images input. Previous NeRF-based segmentation methods have relied on time-consuming neural scene optimization. While recent 3D Gaussian Splatting has notably improved speed, existing Gaussian-based segmentation methods struggle to produce compact masks, especially in zero-shot segmentation. This issue probably stems from their straightforward assignment of learnable parameters to each Gaussian, resulting in a lack of robustness against cross-view inconsistent 2D machine-generated labels. Our method aims to address this problem by employing Dual Feature Fusion Network as Gaussians' segmentation field. Specifically, we first optimize 3D Gaussians under RGB supervision. After Gaussian Locating, DINO features extracted from images are applied through explicit unprojection, which are further incorporated with spatial features from the efficient point cloud processing network. Feature aggregation is utilized to fuse them in a global-to-local strategy for compact segmentation features. Experimental results show that our model outperforms baselines on both semantic and panoptic zero-shot segmentation task, meanwhile consumes less than 10\% inference time compared to NeRF-based methods. Code and more results will be available at //David-Dou.github.io/CoSSegGaussians.
Deep learning has shown great potential for modeling the physical dynamics of complex particle systems such as fluids (in Lagrangian descriptions). Existing approaches, however, require the supervision of consecutive particle properties, including positions and velocities. In this paper, we consider a partially observable scenario known as fluid dynamics grounding, that is, inferring the state transitions and interactions within the fluid particle systems from sequential visual observations of the fluid surface. We propose a differentiable two-stage network named NeuroFluid. Our approach consists of (i) a particle-driven neural renderer, which involves fluid physical properties into the volume rendering function, and (ii) a particle transition model optimized to reduce the differences between the rendered and the observed images. NeuroFluid provides the first solution to unsupervised learning of particle-based fluid dynamics by training these two models jointly. It is shown to reasonably estimate the underlying physics of fluids with different initial shapes, viscosity, and densities. It is a potential alternative approach to understanding complex fluid mechanics, such as turbulence, that are difficult to model using traditional methods of mathematical physics.
Meta reinforcement learning (meta-RL) extracts knowledge from previous tasks and achieves fast adaptation to new tasks. Despite recent progress, efficient exploration in meta-RL remains a key challenge in sparse-reward tasks, as it requires quickly finding informative task-relevant experiences in both meta-training and adaptation. To address this challenge, we explicitly model an exploration policy learning problem for meta-RL, which is separated from exploitation policy learning, and introduce a novel empowerment-driven exploration objective, which aims to maximize information gain for task identification. We derive a corresponding intrinsic reward and develop a new off-policy meta-RL framework, which efficiently learns separate context-aware exploration and exploitation policies by sharing the knowledge of task inference. Experimental evaluation shows that our meta-RL method significantly outperforms state-of-the-art baselines on various sparse-reward MuJoCo locomotion tasks and more complex sparse-reward Meta-World tasks.
Graph representation learning resurges as a trending research subject owing to the widespread use of deep learning for Euclidean data, which inspire various creative designs of neural networks in the non-Euclidean domain, particularly graphs. With the success of these graph neural networks (GNN) in the static setting, we approach further practical scenarios where the graph dynamically evolves. Existing approaches typically resort to node embeddings and use a recurrent neural network (RNN, broadly speaking) to regulate the embeddings and learn the temporal dynamics. These methods require the knowledge of a node in the full time span (including both training and testing) and are less applicable to the frequent change of the node set. In some extreme scenarios, the node sets at different time steps may completely differ. To resolve this challenge, we propose EvolveGCN, which adapts the graph convolutional network (GCN) model along the temporal dimension without resorting to node embeddings. The proposed approach captures the dynamism of the graph sequence through using an RNN to evolve the GCN parameters. Two architectures are considered for the parameter evolution. We evaluate the proposed approach on tasks including link prediction, edge classification, and node classification. The experimental results indicate a generally higher performance of EvolveGCN compared with related approaches. The code is available at \url{//github.com/IBM/EvolveGCN}.
State-of-the-art Convolutional Neural Network (CNN) benefits a lot from multi-task learning (MTL), which learns multiple related tasks simultaneously to obtain shared or mutually related representations for different tasks. The most widely-used MTL CNN structure is based on an empirical or heuristic split on a specific layer (e.g., the last convolutional layer) to minimize different task-specific losses. However, this heuristic sharing/splitting strategy may be harmful to the final performance of one or multiple tasks. In this paper, we propose a novel CNN structure for MTL, which enables automatic feature fusing at every layer. Specifically, we first concatenate features from different tasks according to their channel dimension, and then formulate the feature fusing problem as discriminative dimensionality reduction. We show that this discriminative dimensionality reduction can be done by 1x1 Convolution, Batch Normalization, and Weight Decay in one CNN, which we refer to as Neural Discriminative Dimensionality Reduction (NDDR). We perform ablation analysis in details for different configurations in training the network. The experiments carried out on different network structures and different task sets demonstrate the promising performance and desirable generalizability of our proposed method.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191