This paper investigates the use of deep reinforcement learning (DRL) for the control of mobile robot teams within the context of navigation and task-based collaborative scenarios. We apply a DRL policy with a tailored neural network architecture as a solution to control, path planning, and higher-level guidance tasks. Our network architecture was trained using a unique multi-stage curriculum that progresses from single-agent navigation, to multi-agent pathfinding with obstacles, and finally to a complex collaborative firefighting scenario. This structured approach accelerates training convergence by systematically building sophisticated collaborative behaviours upon foundational skills, which enhances training stability and guides the agents towards learning effective and coordinated strategies The policy evaluation was conducted in both simulation and hybrid simulation-physical demonstrations utilising a real unmanned ground vehicle (UGV). The policy presented is capable of achieving multi-agent navigation tasks with a 95.83% accuracy in our testing environments, and has demonstrated emergent multi-agent behaviours. In more complex collaborative firefighting scenarios, the policy also demonstrated superior performance than baselines in reaching goals, e.g., navigating and extinguishing two fires with a 99.67% success rate, suggesting its strong potential for real-world deployment.

Decentralized online planning, such as decentralized Monte Carlo tree search (MCTS), is an attractive paradigm for cooperative multi-agent systems in information-gathering tasks. However, current MCTS algorithms implicitly assume that agents are always available and actively contributing throughout the mission. In realistic, dynamic, and volatile environments, agent attrition is common and can severely degrade performance. In this paper, we demonstrate that agent attrition can cause current decentralized MCTS methods to perform arbitrarily worse than the optimum, particularly in applications with submodular reward functions. To address this issue, we propose Attritable Monte Carlo Tree Search (A-MCTS), a decentralized MCTS algorithm that adapts quickly and efficiently to reductions in the set of active agents. Our key idea is to have each agent build its search tree using the global utility function while coordinating with teammates through a regret-matching algorithm. Our theoretical analysis and extensive simulations show that A-MCTS maintains effective coordination among agents, even in high-attrition environments. We evaluate our approach in different information-gathering problems by modeling realistic reference scenarios. Results highlight that A-MCTS substantially improves over the main competing methods regarding global utility, scalability, and robustness.

Concept erasure has emerged as a promising technique for mitigating the risk of harmful content generation in diffusion models by selectively unlearning undesirable concepts. The common principle of previous works to remove a specific concept is to map it to a fixed generic concept, such as a neutral concept or just an empty text prompt. In this paper, we demonstrate that this fixed-target strategy is suboptimal, as it fails to account for the impact of erasing one concept on the others. To address this limitation, we model the concept space as a graph and empirically analyze the effects of erasing one concept on the remaining concepts. Our analysis uncovers intriguing geometric properties of the concept space, where the influence of erasing a concept is confined to a local region. Building on this insight, we propose the Adaptive Guided Erasure (AGE) method, which dynamically selects optimal target concepts tailored to each undesirable concept, minimizing unintended side effects. Experimental results show that AGE significantly outperforms state-of-the-art erasure methods on preserving unrelated concepts while maintaining effective erasure performance. Our code is published at {https://github.com/tuananhbui89/Adaptive-Guided-Erasure}.

Coordinated multi-robot systems are an effective way to harvest data from sensor networks and implement active perception strategies. However, achieving efficient coordination in a way that guarantees a target QoS while adapting dynamically to changes (in the environment and/or in the system) is a key open issue. In this paper, we propose a novel decentralized Monte Carlo Tree Search (MCTS) algorithm for dynamic environments that allows agents to optimize their own actions while achieving some form of coordination. Its main underlying idea is to balance adaptively the exploration-exploitation trade-off to deal effectively with changes in the environment while filtering out outdated and irrelevant samples via a sliding window mechanism. We show both theoretically and through simulations that in dynamic environments our algorithm provides a log-factor (in terms of time steps) smaller regret than state-of-the-art decentralized multi-agent planning methods. We instantiate our approach to the problem of underwater data collection, showing in a variety of different settings that our approach greatly outperforms the best-competing approaches, both in terms of convergence speed and global utility.

Coordinated multi-robot systems are an effective way to harvest data from sensor networks and to implement active perception strategies. However, achieving efficient coordination in a way which guarantees a target QoS while adapting dynamically to changes (in the environment, due to sensors’ mobility, and/or in the value of harvested data) is to date a key open issue. In this paper, we propose a novel decentralized Monte Carlo Tree Search algorithm (MCTS) which allows agents to optimize their own actions while achieving some form of coordination, in a changing environment. Its key underlying idea is to balance in an adaptive manner the exploration-exploitation trade-off to deal effectively with abrupt changes caused by the environment and random changes caused by other agents’ actions. Critically, outdated and irrelevant samples - an inherent and prevalent feature in all multi-agent MCTS-based algorithms - are filtered out by means of a sliding window mechanism. We show both theoretically and through simulations that our algorithm provides a log-factor (in terms of time steps) smaller regret than state-of-the-art decentralized multi-agent planning methods. We instantiate our approach on the problem of underwater data collection, showing on a set of different models for changes that our approach greatly outperforms the best available algorithms for that setting, both in terms of convergence speed and of global utility.

Semi-Supervised Support Vector Machines (S3VMs) have been proposed to deal with the proliferation of partially labelled data available in many large-scale complex systems. Since most existing S3VMs do not correspond to convex problems nor have practical solutions especially on large imbalanced data, this limits their utility in practice. In this work, we propose an efficient approach to the semi-supervised problem especially for anomaly detection, termed S3VMAD. We formulate S3VMAD in unsupervised paradigm for anomaly detection. It is a convex objective function which is guaranteed to have a global optimum solution. The proposed approach is compatible with both supervised Support Vector Machines (SVMs) and unsupervised One-Class SVM (OCSVM). In addition, it exploits efficient optimisation techniques in an online fashion as well as in batch mode, which enables the solution to be scalable and therefore practical in the case of large datasets. In order to investigate the efficiency and effectiveness of the proposed S3VMAD, experiments on various sample datasets compare with supervised SVM learning, Semi-Supervised Anomaly Detection (SSAD), and unsupervised learning.

Distance metric learning is of fundamental interest in machine learning because the employed distance metric can significantly affect the performance of many learning methods. Quadratic Mahalanobis metric learning is a popular approach to the problem, but typically requires solving a semidefinite programming (SDP) problem, which is computationally expensive. The worst case complexity of solving an SDP problem involving a matrix variable of size D×D with O(D) linear constraints is about O(D(6.5)) using interior-point methods, where D is the dimension of the input data. Thus, the interior-point methods only practically solve problems exhibiting less than a few thousand variables. Because the number of variables is D(D+1)/2, this implies a limit upon the size of problem that can practically be solved around a few hundred dimensions. The complexity of the popular quadratic Mahalanobis metric learning approach thus limits the size of problem to which metric learning can be applied. Here, we propose a significantly more efficient and scalable approach to the metric learning problem based on the Lagrange dual formulation of the problem. The proposed formulation is much simpler to implement, and therefore allows much larger Mahalanobis metric learning problems to be solved. The time complexity of the proposed method is roughly O(D(3)), which is significantly lower than that of the SDP approach. Experiments on a variety of data sets demonstrate that the proposed method achieves an accuracy comparable with the state of the art, but is applicable to significantly larger problems. We also show that the proposed method can be applied to solve more general Frobenius norm regularized SDP problems approximately.

We concern the problem of learning a Mahalanobis distance metric for improving nearest neighbor classification. Our work is built upon the large margin nearest neighbor (LMNN) classification framework. Due to the semidefiniteness constraint in the optimization problem of LMNN, it is not scalable in terms of the dimensionality of the input data. The original LMNN solver partially alleviates this problem by adopting alternating projection methods instead of standard interior-point methods. Still, at each iteration, the computation complexity is at least O(D3) (D is the dimension of input data). In this work, we propose a column generation based algorithm to solve the LMNN optimization problem much more efficiently. Our algorithm is much more scalable in tha tat each iteration, it does not need full eigen-decomposition. Instead, we only need to find the leading eigen value and its corresponding eigen vector, which is of O(D2) complexity. Experiments show the efficiency and efficacy of our algorithms.

Distance metric learning plays an important role in many vision problems. Previous work of quadratic Mahalanobis metric learning usually needs to solve a semidefinite programming (SDP) problem. A standard interior-point SDP solver has a complexity of O(D6.5) (with D the dimension of input data), and can only solve problems up to a few thousand variables. Since the number of variables is D(D + l)/2, this corresponds to a limit around D3) with t ≈ 20 ~ 30 for most problems in our experiments. The proposed algorithm is scalable and easy to implement. Experiments on various datasets show its similar accuracy compared with state-of-the-art. We also demonstrate that this idea may also be able to be applied to other SDP problems such as maximum variance unfolding.

Kernel-based mean shift (MS) trackers have proven to be a promising alternative to stochastic particle filtering trackers. Despite its popularity, MS trackers have two fundamental drawbacks: 1) the template model can only be built from a single image, and 2) it is difficult to adaptively update the template model. In this paper, we generalize the plain MS trackers and attempt to overcome these two limitations. It is well known that modeling and maintaining a representation of a target object is an important component of a successful visual tracker. However, little work has been done on building a robust template model for kernel-based MS tracking. In contrast to building a template from a single frame, we train a robust object representation model from a large amount of data. Tracking is viewed as a binary classification problem, and a discriminative classification rule is learned to distinguish between the object and background. We adopt a support vector machine for training. The tracker is then implemented by maximizing the classification score. An iterative optimization scheme very similar to MS is derived for this purpose. Compared with the plain MS tracker, it is now much easier to incorporate online template adaptation to cope with inherent changes during the course of tracking. To this end, a sophisticated online support vector machine is used. We demonstrate successful localization and tracking on various data sets.

For many machine learning algorithms such as k-nearest neighbor ( k-NN) classifiers and k-means clustering, often their success heavily depends on the metric used to calculate distances between different data points. An effective solution for defining such a metric is to learn it from a set of labeled training samples. In this work, we propose a fast and scalable algorithm to learn a Mahalanobis distance metric. The Mahalanobis metric can be viewed as the Euclidean distance metric on the input data that have been linearly transformed. By employing the principle of margin maximization to achieve better generalization performances, this algorithm formulates the metric learning as a convex optimization problem and a positive semidefinite (p.s.d.) matrix is the unknown variable. Based on an important theorem that a p.s.d. trace-one matrix can always be represented as a convex combination of multiple rank-one matrices, our algorithm accommodates any differentiable loss function and solves the resulting optimization problem using a specialized gradient descent procedure. During the course of optimization, the proposed algorithm maintains the positive semidefiniteness of the matrix variable that is essential for a Mahalanobis metric. Compared with conventional methods like standard interior-point algorithms or the special solver used in large margin nearest neighbor , our algorithm is much more efficient and has a better performance in scalability. Experiments on benchmark data sets suggest that, compared with state-of-the-art metric learning algorithms, our algorithm can achieve a comparable classification accuracy with reduced computational complexity.

The learning of appropriate distance metrics is a critical problem in classification. In this work, we propose a boosting-based technique, termed BoostMetric, for learning a Mahalanobis distance metric. One of the primary difficulties in learning such a metric is to ensure that the Mahalanobis matrix remains positive semidefinite. Semidefinite programming is sometimes used to enforce this constraint, but does not scale well. BoostMetric is instead based on a key observation that any positive semidefinite matrix can be decomposed into a linear positive combination of trace-one rank-one matrices. BoostMetric thus uses rank-one positive semidefinite matrices as weak learners within an efficient and scalable boosting-based learning process. The resulting method is easy to implement, does not require tuning, and can accommodate various types of constraints. Experiments on various datasets show that the proposed algorithm compares favorably to those state-of-the-art methods in terms of classification accuracy and running time.