Regret minimization methods are a powerful tool for learning approximate Nash equilibrium in two-player zero-sum imperfect information extensive-form games (IIEGs). We consider the problem in the interactive bandit-feedback setting where we don't know the dynamics of the IIEG. In general, only the interactive trajectory and the loss $(\ell^t)^Tx^t$ are revealed. To learn approximate Nash equilibrium, the regret minimizer is required to estimate the full-feedback loss gradient $\ell^t$ and minimize the regret. In this paper, we propose a generalized framework for this learning setting. We demonstrate that the most recent bandit regret minimization methods, including MCCFR, IXOMD, and Balanced OMD, can be analyzed as a particular case of our framework. It presents a theoretical framework for the design and the modular analysis of the bandit regret minimization methods. Precisely, it allows us to use any gradient estimator, any exploration strategy, any sampling strategy, coupled with any full-feedback regret minimization methods.
Although deep reinforcement learning has become a universal solution for complex control tasks, its real-world applicability is still limited because lacking security guarantees for policies. To address this problem, we propose Boundary Characterization via the Minimum Experience Retention (BCMER), an end-to-end Interpretable Policy Distillation (IPD) framework. Unlike previous IPD approaches, BCMER distinguishes the importance of experiences and keeps a minimal but critical experience pool with almost no loss of policy similarity. Specifically, the proposed BCMER contains two basic steps. Firstly, we propose a novel multidimensional hyperspheres intersection (MHI) approach to divide experience points into boundary points and internal points, and reserve the crucial boundary points. Secondly, we develop a nearest-neighbor-based model to generate robust and interpretable decision rules based on the boundary points. Extensive experiments show that the proposed BCMER is able to reduce the amount of experience to 1.4%~19.1% (when the count of the naive experiences is 10k) and maintain high IPD performance. In general, the proposed BCMER is more suitable for the experience storage limited regime because it discovers the critical experience and eliminates redundant experience.
Transformers have dominated the field of natural language processing, and recently impacted the computer vision area. In the field of medical image analysis, Transformers have also been successfully applied to full-stack clinical applications, including image synthesis/reconstruction, registration, segmentation, detection, and diagnosis. Our paper presents both a position paper and a primer, promoting awareness and application of Transformers in the field of medical image analysis. Specifically, we first overview the core concepts of the attention mechanism built into Transformers and other basic components. Second, we give a new taxonomy of various Transformer architectures tailored for medical image applications and discuss their limitations. Within this review, we investigate key challenges revolving around the use of Transformers in different learning paradigms, improving the model efficiency, and their coupling with other techniques. We hope this review can give a comprehensive picture of Transformers to the readers in the field of medical image analysis.
With rising uncertainty in the real world, online Reinforcement Learning (RL) has been receiving increasing attention due to its fast learning capability and improving data efficiency. However, online RL often suffers from complex Value Function Approximation (VFA) and catastrophic interference, creating difficulty for the deep neural network to be applied to an online RL algorithm in a fully online setting. Therefore, a simpler and more adaptive approach is introduced to evaluate value function with the kernel-based model. Sparse representations are superior at handling interference, indicating that competitive sparse representations should be learnable, non-prior, non-truncated and explicit when compared with current sparse representation methods. Moreover, in learning sparse representations, attention mechanisms are utilized to represent the degree of sparsification, and a smooth attentive function is introduced into the kernel-based VFA. In this paper, we propose an Online Attentive Kernel-Based Temporal Difference (OAKTD) algorithm using two-timescale optimization and provide convergence analysis of our proposed algorithm. Experimental evaluations showed that OAKTD outperformed several Online Kernel-based Temporal Difference (OKTD) learning algorithms in addition to the Temporal Difference (TD) learning algorithm with Tile Coding on public Mountain Car, Acrobot, CartPole and Puddle World tasks.
Aiming to generalize the model trained in source domains to unseen target domains, domain generalization (DG) has attracted lots of attention recently. The key issue of DG is how to prevent overfitting to the observed source domains because target domain is unavailable during training. We investigate that overfitting not only causes the inferior generalization ability to unseen target domains but also leads unstable prediction in the test stage. In this paper, we observe that both sampling multiple tasks in training stage and generating augmented images in test stage largely benefit generalization performance. Thus, by treating tasks and images as different views, we propose a novel multi-view DG framework. Specifically, in training stage, to enhance generalization ability, we develop a multi-view regularized meta-learning algorithm that employs multiple tasks to produce a suitable optimization direction during updating model. In test stage, to alleviate unstable prediction, we utilize multiple augmented images to yield multi-view prediction, which significantly promotes model reliability via fusing the results of different views of a test image. Extensive experiments on three benchmark datasets validate our method outperforms several state-of-the-art approaches.
An AI agent should be able to coordinate with humans to solve tasks. We consider the problem of training a Reinforcement Learning (RL) agent without using any human data, i.e., in a zero-shot setting, to make it capable of collaborating with humans. Standard RL agents learn through self-play. Unfortunately, these agents only know how to collaborate with themselves and normally do not perform well with unseen partners, such as humans. The methodology of how to train a robust agent in a zero-shot fashion is still subject to research. Motivated from the maximum entropy RL, we derive a centralized population entropy objective to facilitate learning of a diverse population of agents, which is later used to train a robust agent to collaborate with unseen partners. The proposed method shows its effectiveness compared to baseline methods, including self-play PPO, the standard Population-Based Training (PBT), and trajectory diversity-based PBT, in the popular Overcooked game environment. We also conduct online experiments with real humans and further demonstrate the efficacy of the method in the real world. A supplementary video showing experimental results is available at https://youtu.be/Xh-FKD0AAKE.
By training a model on multiple observed source domains, domain generalization aims to generalize well to arbitrary unseen target domains without further training. Existing works mainly focus on learning domain-invariant features to improve the generalization ability. However, since target domain is not available during training, previous methods inevitably suffer from overfitting in source domains. To tackle this issue, we develop an effective dropout-based framework to enlarge the region of the model's attention, which can effectively mitigate the overfitting problem. Particularly, different from the typical dropout scheme, which normally conducts the dropout on the fixed layer, first, we randomly select one layer, and then we randomly select its channels to conduct dropout. Besides, we leverage the progressive scheme to add the ratio of the dropout during training, which can gradually boost the difficulty of training model to enhance the robustness of the model. Moreover, to further alleviate the impact of the overfitting issue, we leverage the augmentation schemes on image-level and feature-level to yield a strong baseline model. We conduct extensive experiments on multiple benchmark datasets, which show our method can outperform the state-of-the-art methods.
Efficient exploration in deep cooperative multi-agent reinforcement learning (MARL) still remains challenging in complex coordination problems. In this paper, we introduce a novel Episodic Multi-agent reinforcement learning with Curiosity-driven exploration, called EMC. We leverage an insight of popular factorized MARL algorithms that the "induced" individual Q-values, i.e., the individual utility functions used for local execution, are the embeddings of local action-observation histories, and can capture the interaction between agents due to reward backpropagation during centralized training. Therefore, we use prediction errors of individual Q-values as intrinsic rewards for coordinated exploration and utilize episodic memory to exploit explored informative experience to boost policy training. As the dynamics of an agent's individual Q-value function captures the novelty of states and the influence from other agents, our intrinsic reward can induce coordinated exploration to new or promising states. We illustrate the advantages of our method by didactic examples, and demonstrate its significant outperformance over state-of-the-art MARL baselines on challenging tasks in the StarCraft II micromanagement benchmark.
Quantitative susceptibility mapping (QSM) is a valuable MRI post-processing technique that quantifies the magnetic susceptibility of body tissue from phase data. However, the traditional QSM reconstruction pipeline involves multiple non-trivial steps, including phase unwrapping, background field removal, and dipole inversion. These intermediate steps not only increase the reconstruction time but amplify noise and errors. This study develops a large-stencil Laplacian preprocessed deep learning-based neural network for near instant quantitative field and susceptibility mapping (i.e., iQFM and iQSM) from raw MR phase data. The proposed iQFM and iQSM methods were compared with established reconstruction pipelines on simulated and in vivo datasets. In addition, experiments on patients with intracranial hemorrhage and multiple sclerosis were also performed to test the generalization of the novel neural networks. The proposed iQFM and iQSM methods yielded comparable results to multi-step methods in healthy subjects while dramatically improving reconstruction accuracies on intracranial hemorrhages with large susceptibilities. The reconstruction time was also substantially shortened from minutes using multi-step methods to only 30 milliseconds using the trained iQFM and iQSM neural networks.
Reinforcement learning has achieved great success in many applications. However, sample efficiency remains a key challenge, with prominent methods requiring millions (or even billions) of environment steps to train. Recently, there has been significant progress in sample efficient image-based RL algorithms; however, consistent human-level performance on the Atari game benchmark remains an elusive goal. We propose a sample efficient model-based visual RL algorithm built on MuZero, which we name EfficientZero. Our method achieves 190.4% mean human performance and 116.0% median performance on the Atari 100k benchmark with only two hours of real-time game experience and outperforms the state SAC in some tasks on the DMControl 100k benchmark. This is the first time an algorithm achieves super-human performance on Atari games with such little data. EfficientZero's performance is also close to DQN's performance at 200 million frames while we consume 500 times less data. EfficientZero's low sample complexity and high performance can bring RL closer to real-world applicability. We implement our algorithm in an easy-to-understand manner and it is available at https://github.com/YeWR/EfficientZero. We hope it will accelerate the research of MCTS-based RL algorithms in the wider community.