Cooperative artificial intelligence with human or superhuman proficiency in collaborative tasks stands at the frontier of machine learning research. Prior work has tended to evaluate cooperative AI performance under the restrictive paradigms of self-play (teams composed of agents trained together) and cross-play (teams of agents trained independently but using the same algorithm). Recent work has indicated that AI optimized for these narrow settings may make for undesirable collaborators in the real-world. We formalize an alternative criteria for evaluating cooperative AI, referred to as inter-algorithm cross-play, where agents are evaluated on teaming performance with all other agents within an experiment pool with no assumption of algorithmic similarities between agents. We show that existing state-of-the-art cooperative AI algorithms, such as Other-Play and Off-Belief Learning, under-perform in this paradigm. We propose the Any-Play learning augmentation -- a multi-agent extension of diversity-based intrinsic rewards for zero-shot coordination (ZSC) -- for generalizing self-play-based algorithms to the inter-algorithm cross-play setting. We apply the Any-Play learning augmentation to the Simplified Action Decoder (SAD) and demonstrate state-of-the-art performance in the collaborative card game Hanabi.
相關內容
Recently, model-based agents have achieved better performance compared with model-free ones using the same computational budget and training time in single-agent environments. However, due to the complexity of multi-agent systems, it is very difficult to learn the model of the environment. When model-based methods are applied to multi-agent tasks, the significant compounding error may hinder the learning process. In this paper, we propose an implicit model-based multi-agent reinforcement learning method based on value decomposition methods. Under this method, agents can interact with the learned virtual environment and evaluate the current state value according to imagined future states, which makes agents have foresight. Our method can be applied to any multi-agent value decomposition method. The experimental results show that our method improves the sample efficiency in partially observable Markov decision process domains.
We consider the problem of training a classification model with group annotated training data. Recent work has established that, if there is distribution shift across different groups, models trained using the standard empirical risk minimization (ERM) objective suffer from poor performance on minority groups and that group distributionally robust optimization (Group-DRO) objective is a better alternative. The starting point of this paper is the observation that though Group-DRO performs better than ERM on minority groups for some benchmark datasets, there are several other datasets where it performs much worse than ERM. Inspired by ideas from the closely related problem of domain generalization, this paper proposes a new and simple algorithm that explicitly encourages learning of features that are shared across various groups. The key insight behind our proposed algorithm is that while Group-DRO focuses on groups with worst regularized loss, focusing instead, on groups that enable better performance even on other groups, could lead to learning of shared/common features, thereby enhancing minority performance beyond what is achieved by Group-DRO. Empirically, we show that our proposed algorithm matches or achieves better performance compared to strong contemporary baselines including ERM and Group-DRO on standard benchmarks on both minority groups and across all groups. Theoretically, we show that the proposed algorithm is a descent method and finds first order stationary points of smooth nonconvex functions.
Approximately 50% of development resources are devoted to UI development tasks [9]. Occupying a large proportion of development resources, developing icons can be a time-consuming task, because developers need to consider not only effective implementation methods but also easy-to-understand descriptions. In this paper, we present Auto-Icon+, an approach for automatically generating readable and efficient code for icons from design artifacts. According to our interviews to understand the gap between designers (icons are assembled from multiple components) and developers (icons as single images), we apply a heuristic clustering algorithm to compose the components into an icon image. We then propose an approach based on a deep learning model and computer vision methods to convert the composed icon image to fonts with descriptive labels, thereby reducing the laborious manual effort for developers and facilitating UI development. We quantitatively evaluate the quality of our method in the real world UI development environment and demonstrate that our method offers developers accurate, efficient, readable, and usable code for icon designs, in terms of saving 65.2% implementing time.
Embodied AI is a recent research area that aims at creating intelligent agents that can move and operate inside an environment. Existing approaches in this field demand the agents to act in completely new and unexplored scenes. However, this setting is far from realistic use cases that instead require executing multiple tasks in the same environment. Even if the environment changes over time, the agent could still count on its global knowledge about the scene while trying to adapt its internal representation to the current state of the environment. To make a step towards this setting, we propose Spot the Difference: a novel task for Embodied AI where the agent has access to an outdated map of the environment and needs to recover the correct layout in a fixed time budget. To this end, we collect a new dataset of occupancy maps starting from existing datasets of 3D spaces and generating a number of possible layouts for a single environment. This dataset can be employed in the popular Habitat simulator and is fully compliant with existing methods that employ reconstructed occupancy maps during navigation. Furthermore, we propose an exploration policy that can take advantage of previous knowledge of the environment and identify changes in the scene faster and more effectively than existing agents. Experimental results show that the proposed architecture outperforms existing state-of-the-art models for exploration on this new setting.
The dynamic response of the legged robot locomotion is non-Lipschitz and can be stochastic due to environmental uncertainties. To test, validate, and characterize the safety performance of legged robots, existing solutions on observed and inferred risk can be incomplete and sampling inefficient. Some formal verification methods suffer from the model precision and other surrogate assumptions. In this paper, we propose a scenario sampling based testing framework that characterizes the overall safety performance of a legged robot by specifying (i) where (in terms of a set of states) the robot is potentially safe, and (ii) how safe the robot is within the specified set. The framework can also help certify the commercial deployment of the legged robot in real-world environment along with human and compare safety performance among legged robots with different mechanical structures and dynamic properties. The proposed framework is further deployed to evaluate a group of state-of-the-art legged robot locomotion controllers from various model-based, deep neural network involved, and reinforcement learning based methods in the literature. Among a series of intended work domains of the studied legged robots (e.g. tracking speed on sloped surface, with abrupt changes on demanded velocity, and against adversarial push-over disturbances), we show that the method can adequately capture the overall safety characterization and the subtle performance insights. Many of the observed safety outcomes, to the best of our knowledge, have never been reported by the existing work in the legged robot literature.
Designers reportedly struggle with design optimization tasks where they are asked to find a combination of design parameters that maximizes a given set of objectives. In HCI, design optimization problems are often exceedingly complex, involving multiple objectives and expensive empirical evaluations. Model-based computational design algorithms assist designers by generating design examples during design, however they assume a model of the interaction domain. Black box methods for assistance, on the other hand, can work with any design problem. However, virtually all empirical studies of this human-in-the-loop approach have been carried out by either researchers or end-users. The question stands out if such methods can help designers in realistic tasks. In this paper, we study Bayesian optimization as an algorithmic method to guide the design optimization process. It operates by proposing to a designer which design candidate to try next, given previous observations. We report observations from a comparative study with 40 novice designers who were tasked to optimize a complex 3D touch interaction technique. The optimizer helped designers explore larger proportions of the design space and arrive at a better solution, however they reported lower agency and expressiveness. Designers guided by an optimizer reported lower mental effort but also felt less creative and less in charge of the progress. We conclude that human-in-the-loop optimization can support novice designers in cases where agency is not critical.
Deep learning depends on large amounts of labeled training data. Manual labeling is expensive and represents a bottleneck, especially for tasks such as segmentation, where labels must be assigned down to the level of individual points. That challenge is even more daunting for 3D data: 3D point clouds contain millions of points per scene, and their accurate annotation is markedly more time-consuming. The situation is further aggravated by the added complexity of user interfaces for 3D point clouds, which slows down annotation even more. For the case of 2D image segmentation, interactive techniques have become common, where user feedback in the form of a few clicks guides a segmentation algorithm -- nowadays usually a neural network -- to achieve an accurate labeling with minimal effort. Surprisingly, interactive segmentation of 3D scenes has not been explored much. Previous work has attempted to obtain accurate 3D segmentation masks using human feedback from the 2D domain, which is only possible if correctly aligned images are available together with the 3D point cloud, and it involves switching between the 2D and 3D domains. Here, we present an interactive 3D object segmentation method in which the user interacts directly with the 3D point cloud. Importantly, our model does not require training data from the target domain: when trained on ScanNet, it performs well on several other datasets with different data characteristics as well as different object classes. Moreover, our method is orthogonal to supervised (instance) segmentation methods and can be combined with them to refine automatic segmentations with minimal human effort.
Proactive dialogue system is able to lead the conversation to a goal topic and has advantaged potential in bargain, persuasion and negotiation. Current corpus-based learning manner limits its practical application in real-world scenarios. To this end, we contribute to advance the study of the proactive dialogue policy to a more natural and challenging setting, i.e., interacting dynamically with users. Further, we call attention to the non-cooperative user behavior -- the user talks about off-path topics when he/she is not satisfied with the previous topics introduced by the agent. We argue that the targets of reaching the goal topic quickly and maintaining a high user satisfaction are not always converge, because the topics close to the goal and the topics user preferred may not be the same. Towards this issue, we propose a new solution named I-Pro that can learn Proactive policy in the Interactive setting. Specifically, we learn the trade-off via a learned goal weight, which consists of four factors (dialogue turn, goal completion difficulty, user satisfaction estimation, and cooperative degree). The experimental results demonstrate I-Pro significantly outperforms baselines in terms of effectiveness and interpretability.
Few-shot Learning aims to learn classifiers for new classes with only a few training examples per class. Existing meta-learning or metric-learning based few-shot learning approaches are limited in handling diverse domains with various number of labels. The meta-learning approaches train a meta learner to predict weights of homogeneous-structured task-specific networks, requiring a uniform number of classes across tasks. The metric-learning approaches learn one task-invariant metric for all the tasks, and they fail if the tasks diverge. We propose to deal with these limitations with meta metric learning. Our meta metric learning approach consists of task-specific learners, that exploit metric learning to handle flexible labels, and a meta learner, that discovers good parameters and gradient decent to specify the metrics in task-specific learners. Thus the proposed model is able to handle unbalanced classes as well as to generate task-specific metrics. We test our approach in the `$k$-shot $N$-way' few-shot learning setting used in previous work and new realistic few-shot setting with diverse multi-domain tasks and flexible label numbers. Experiments show that our approach attains superior performances in both settings.
In this paper, we focus on three problems in deep learning based medical image segmentation. Firstly, U-net, as a popular model for medical image segmentation, is difficult to train when convolutional layers increase even though a deeper network usually has a better generalization ability because of more learnable parameters. Secondly, the exponential ReLU (ELU), as an alternative of ReLU, is not much different from ReLU when the network of interest gets deep. Thirdly, the Dice loss, as one of the pervasive loss functions for medical image segmentation, is not effective when the prediction is close to ground truth and will cause oscillation during training. To address the aforementioned three problems, we propose and validate a deeper network that can fit medical image datasets that are usually small in the sample size. Meanwhile, we propose a new loss function to accelerate the learning process and a combination of different activation functions to improve the network performance. Our experimental results suggest that our network is comparable or superior to state-of-the-art methods.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191