Search for: All records

To mitigate the limitation that the classical reinforcement learning (RL) framework heavily relies on identical training and test environments, Distributionally Robust RL (DRRL) has been proposed to enhance performance across a range of environments, possibly including unknown test environments. As a price for robustness gain, DRRL involves optimizing over a set of distributions, which is inherently more challenging than optimizing over a fixed distribution in the non-robust case. Existing DRRL algorithms are either model-based or fail to learn from a single sample trajectory. In this paper, we design a first fully model-free DRRL algorithm, called distributionally robust Q-learning with single trajectory (DRQ). We delicately design a multi-timescale framework to fully utilize each incrementally arriving sample and directly learn the optimal distributionally robust policy without modeling the environment, thus the algorithm can be trained along a single trajectory in a model-free fashion. Despite the algorithm’s complexity, we provide asymptotic convergence guarantees by generalizing classical stochastic approximation tools. Comprehensive experimental results demonstrate the superior robustness and sample complexity of our proposed algorithm, compared to non-robust methods and other robust RL algorithms. 

To mitigate the limitation that the classical reinforcement learning (RL) framework heavily relies on identical training and test environments, Distributionally Robust RL (DRRL) has been proposed to enhance performance across a range of environments, possibly including unknown test environments. As a price for robustness gain, DRRL involves optimizing over a set of distributions, which is inherently more challenging than optimizing over a fixed distribution in the non-robust case. Existing DRRL algorithms are either model-based or fail to learn from a single sample trajectory. In this paper, we design a first fully model-free DRRL algorithm, called distributionally robust Q-learning with single trajectory (DRQ). We delicately design a multi-timescale framework to fully utilize each incrementally arriving sample and directly learn the optimal distributionally robust policy without modeling the environment, thus the algorithm can be trained along a single trajectory in a model-free fashion. Despite the algorithm's complexity, we provide asymptotic convergence guarantees by generalizing classical stochastic approximation tools.Comprehensive experimental results demonstrate the superior robustness and sample complexity of our proposed algorithm, compared to non-robust methods and other robust RL algorithms. @inproceedings{ liang2024singletrajectory, title={Single-Trajectory Distributionally Robust Reinforcement Learning}, author={Zhipeng Liang and Xiaoteng Ma and Jose Blanchet and Jun Yang and Jiheng Zhang and Zhengyuan Zhou}, booktitle={Forty-first International Conference on Machine Learning}, year={2024}, url={https://openreview.net/forum?id=3B6vmW2L80} } 

To mitigate the limitation that the classical reinforcement learning (RL) framework heavily relies on identical training and test environments, Distributionally Robust RL (DRRL) has been proposed to enhance performance across a range of environments, possibly including unknown test environments. As a price for robustness gain, DRRL involves optimizing over a set of distributions, which is inherently more challenging than optimizing over a fixed distribution in the non-robust case. Existing DRRL algorithms are either model-based or fail to learn from a single sample trajectory. In this paper, we design a first fully model-free DRRL algorithm, called distributionally robust Q-learning with single trajectory (DRQ). We delicately design a multi-timescale framework to fully utilize each incrementally arriving sample and directly learn the optimal distributionally robust policy without modeling the environment, thus the algorithm can be trained along a single trajectory in a model-free fashion. Despite the algorithm's complexity, we provide asymptotic convergence guarantees by generalizing classical stochastic approximation tools. Comprehensive experimental results demonstrate the superior robustness and sample complexity of our proposed algorithm, compared to non-robust methods and other robust RL algorithms. 

Dioecy (separate sexes) has independently evolved numerous times across the angiosperm phylogeny and is recently derived in many lineages. However, our understanding is limited regarding the evolutionary mechanisms that drive the origins of dioecy in plants. The recent and repeated evolution of dioecy across angiosperms offers an opportunity to make strong inferences about the ecological, developmental, and molecular factors influencing the evolution of dioecy, and thus sex chromosomes. The genusAsparagus(Asparagaceae) is an emerging model taxon for studying dioecy and sex chromosome evolution, yet estimates for the age and origin of dioecy in the genus are lacking.

We use plastome sequences and fossil time calibrations in phylogenetic analyses to investigate the age and origin of dioecy in the genusAsparagus. We also review the diversity of sexual systems present across the genus to address contradicting reports in the literature.

We estimate that dioecy evolved once or twice approximately 2.78−3.78 million years ago inAsparagus, of which roughly 27% of the species are dioecious and the remaining are hermaphroditic with monoclinous flowers.

Our findings support previous work implicating a young age and the possibility of two origins of dioecy inAsparagus, which appear to be associated with rapid radiations and range expansion out of Africa. Lastly, we speculate that paleoclimatic oscillations throughout northern Africa may have helped set the stage for the origin(s) of dioecy inAsparagusapproximately 2.78−3.78 million years ago.

 

The construction of a better exchange-correlation potential in time-dependent density functional theory (TDDFT) can improve the accuracy of TDDFT calculations and provide more accurate predictions of the properties of many-electron systems. Here, we propose a machine learning method to develop the energy functional and the Kohn–Sham potential of a time-dependent Kohn–Sham (TDKS) system is proposed. The method is based on the dynamics of the Kohn–Sham system and does not require any data on the exact Kohn–Sham potential for training the model. We demonstrate the results of our method with a 1D harmonic oscillator example and a 1D two-electron example. We show that the machine-learned Kohn–Sham potential matches the exact Kohn–Sham potential in the absence of memory effect. Our method can still capture the dynamics of the Kohn–Sham system in the presence of memory effects. The machine learning method developed in this article provides insight into making better approximations of the energy functional and the Kohn–Sham potential in the TDKS system.