Models, code, and papers for "Longbo Huang":

Multi-armed Bandits with Compensation

Nov 05, 2018
Siwei Wang, Longbo Huang

We propose and study the known-compensation multi-arm bandit (KCMAB) problem, where a system controller offers a set of arms to many short-term players for $T$ steps. In each step, one short-term player arrives to the system. Upon arrival, the player aims to select an arm with the current best average reward and receives a stochastic reward associated with the arm. In order to incentivize players to explore other arms, the controller provides a proper payment compensation to players. The objective of the controller is to maximize the total reward collected by players while minimizing the compensation. We first provide a compensation lower bound $\Theta(\sum_i {\Delta_i\log T\over KL_i})$, where $\Delta_i$ and $KL_i$ are the expected reward gap and Kullback-Leibler (KL) divergence between distributions of arm $i$ and the best arm, respectively. We then analyze three algorithms to solve the KCMAB problem, and obtain their regrets and compensations. We show that the algorithms all achieve $O(\log T)$ regret and $O(\log T)$ compensation that match the theoretical lower bound. Finally, we present experimental results to demonstrate the performance of the algorithms.

  Click for Model/Code and Paper
Exploring Fast Algorithms for Composite Optimization with Serial and Asynchronous Realizations

Sep 06, 2018
Yue Yu, Longbo Huang

We provide a comprehensive analysis of stochastic variance reduced gradient (SVRG) based proximal algorithms, both with and without momentum, in serial and asynchronous realizations. Specifically, we propose the Prox-SVRG$^{++}$ algorithm, and prove that it has a linear convergence rate with a smaller epoch length (than condition number). Then, we propose a momentum accelerated algorithm, called Prox-MSVRG$^{++}$, and show that it achieves a complexity of $O(\frac{1}{\sqrt{\epsilon}})$. After that, we develop two asynchronous versions of the above serial algorithms and provide a general analysis under nonconvex and non-strongly convex cases respectively. Our theoretical results indicate that the algorithms can achieve a significant speedup when implemented with multiple servers. We conduct extensive experiments based on $4$ real-world datasets on an experiment platform with $11$ physical machines. The experiments validate our theoretical findings and demonstrate the effectiveness of our algorithms.

  Click for Model/Code and Paper
Fast Stochastic Variance Reduced ADMM for Stochastic Composition Optimization

May 22, 2017
Yue Yu, Longbo Huang

We consider the stochastic composition optimization problem proposed in \cite{wang2017stochastic}, which has applications ranging from estimation to statistical and machine learning. We propose the first ADMM-based algorithm named com-SVR-ADMM, and show that com-SVR-ADMM converges linearly for strongly convex and Lipschitz smooth objectives, and has a convergence rate of $O( \log S/S)$, which improves upon the $O(S^{-4/9})$ rate in \cite{wang2016accelerating} when the objective is convex and Lipschitz smooth. Moreover, com-SVR-ADMM possesses a rate of $O(1/\sqrt{S})$ when the objective is convex but without Lipschitz smoothness. We also conduct experiments and show that it outperforms existing algorithms.

  Click for Model/Code and Paper
Multi-Path Policy Optimization

Nov 22, 2019
Ling Pan, Qingpeng Cai, Longbo Huang

Recent years have witnessed a tremendous improvement of deep reinforcement learning. However, a challenging problem is that an agent may suffer from inefficient exploration, particularly for on-policy methods. Previous exploration methods either rely on complex structure to estimate the novelty of states, or incur sensitive hyper-parameters causing instability. In this paper, we propose an efficient exploration method, Multi-Path Policy Optimization (MPPO), which does not incur high computation cost and ensures stability. MPPO maintains an efficient mechanism that effectively utilizes a population of diverse policies to enable better exploration, especially in sparse environments. We also give a theoretical guarantee of the stable performance. We build our scheme upon two widely-adopted on-policy methods, the Trust-Region Policy Optimization (TRPO) algorithm and Proximal Policy Optimization (PPO) algorithm. We conduct extensive experiments on several MuJoCo tasks and their sparsified variants to fairly evaluate the proposed method. Results show that MPPO significantly outperforms state-of-the-art exploration methods in terms of both sample efficiency and final performance.

  Click for Model/Code and Paper
The Power of Online Learning in Stochastic Network Optimization

Jul 29, 2014
Longbo Huang, Xin Liu, Xiaohong Hao

In this paper, we investigate the power of online learning in stochastic network optimization with unknown system statistics {\it a priori}. We are interested in understanding how information and learning can be efficiently incorporated into system control techniques, and what are the fundamental benefits of doing so. We propose two \emph{Online Learning-Aided Control} techniques, $\mathtt{OLAC}$ and $\mathtt{OLAC2}$, that explicitly utilize the past system information in current system control via a learning procedure called \emph{dual learning}. We prove strong performance guarantees of the proposed algorithms: $\mathtt{OLAC}$ and $\mathtt{OLAC2}$ achieve the near-optimal $[O(\epsilon), O([\log(1/\epsilon)]^2)]$ utility-delay tradeoff and $\mathtt{OLAC2}$ possesses an $O(\epsilon^{-2/3})$ convergence time. $\mathtt{OLAC}$ and $\mathtt{OLAC2}$ are probably the first algorithms that simultaneously possess explicit near-optimal delay guarantee and sub-linear convergence time. Simulation results also confirm the superior performance of the proposed algorithms in practice. To the best of our knowledge, our attempt is the first to explicitly incorporate online learning into stochastic network optimization and to demonstrate its power in both theory and practice.

  Click for Model/Code and Paper
Rebalancing Dockless Bike Sharing Systems

Sep 10, 2018
Ling Pan, Qingpeng Cai, Zhixuan Fang, Pingzhong Tang, Longbo Huang

Bike sharing provides an environment-friendly way for traveling and is booming all over the world. Yet, due to the high similarity of user travel patterns, the bike imbalance problem constantly occurs, especially for dockless bike sharing systems, causing significant impact on service quality and company revenue. Thus, it has become a critical task for bike sharing systems to resolve such imbalance efficiently. In this paper, we propose a novel deep reinforcement learning framework for incentivizing users to rebalance such systems. We model the problem as a Markov decision process and take both spatial and temporal features into consideration. We develop a novel deep reinforcement learning algorithm called Hierarchical Reinforcement Pricing (HRP), which builds upon the Deep Deterministic Policy Gradient algorithm. Different from existing methods that often ignore spatial information and rely heavily on accurate prediction, HRP captures both spatial and temporal dependencies using a divide-and-conquer structure with an embedded localized module. We conduct extensive experiments to evaluate HRP, based on a dataset from Mobike, a major Chinese dockless bike sharing company. Results show that HRP performs close to the 24-timeslot look-ahead optimization, and outperforms state-of-the-art methods in both service level and bike distribution. It also transfers well when applied to unseen areas.

  Click for Model/Code and Paper
Beyond the Click-Through Rate: Web Link Selection with Multi-level Feedback

May 04, 2018
Kun Chen, Kechao Cai, Longbo Huang, John C. S. Lui

The web link selection problem is to select a small subset of web links from a large web link pool, and to place the selected links on a web page that can only accommodate a limited number of links, e.g., advertisements, recommendations, or news feeds. Despite the long concerned click-through rate which reflects the attractiveness of the link itself, the revenue can only be obtained from user actions after clicks, e.g., purchasing after being directed to the product pages by recommendation links. Thus, the web links have an intrinsic \emph{multi-level feedback structure}. With this observation, we consider the context-free web link selection problem, where the objective is to maximize revenue while ensuring that the attractiveness is no less than a preset threshold. The key challenge of the problem is that each link's multi-level feedbacks are stochastic, and unobservable unless the link is selected. We model this problem with a constrained stochastic multi-armed bandit formulation, and design an efficient link selection algorithm, called Constrained Upper Confidence Bound algorithm (\textbf{Con-UCB}), and prove $O(\sqrt{T\ln T})$ bounds on both the regret and the violation of the attractiveness constraint. We conduct extensive experiments on three real-world datasets, and show that \textbf{Con-UCB} outperforms state-of-the-art context-free bandit algorithms concerning the multi-level feedback structure.

* 8 pages, 12 figures 

  Click for Model/Code and Paper
Multi-level Feedback Web Links Selection Problem: Learning and Optimization

Sep 08, 2017
Kechao Cai, Kun Chen, Longbo Huang, John C. S. Lui

Selecting the right web links for a website is important because appropriate links not only can provide high attractiveness but can also increase the website's revenue. In this work, we first show that web links have an intrinsic \emph{multi-level feedback structure}. For example, consider a $2$-level feedback web link: the $1$st level feedback provides the Click-Through Rate (CTR) and the $2$nd level feedback provides the potential revenue, which collectively produce the compound $2$-level revenue. We consider the context-free links selection problem of selecting links for a homepage so as to maximize the total compound $2$-level revenue while keeping the total $1$st level feedback above a preset threshold. We further generalize the problem to links with $n~(n\ge2)$-level feedback structure. The key challenge is that the links' multi-level feedback structures are unobservable unless the links are selected on the homepage. To our best knowledge, we are the first to model the links selection problem as a constrained multi-armed bandit problem and design an effective links selection algorithm by learning the links' multi-level structure with provable \emph{sub-linear} regret and violation bounds. We uncover the multi-level feedback structures of web links in two real-world datasets. We also conduct extensive experiments on the datasets to compare our proposed \textbf{LExp} algorithm with two state-of-the-art context-free bandit algorithms and show that \textbf{LExp} algorithm is the most effective in links selection while satisfying the constraint.

* 8 pages (with full proof), 4 figures, ICDM 2017 technical report 

  Click for Model/Code and Paper
Reinforcement Learning with Dynamic Boltzmann Softmax Updates

Mar 15, 2019
Ling Pan, Qingpeng Cai, Qi Meng, Wei Chen, Longbo Huang, Tie-Yan Liu

Value function estimation is an important task in reinforcement learning, i.e., prediction. The commonly used operator for prediction in Q-learning is the hard max operator, which always commits to the maximum action-value according to current estimation. Such `hard' updating scheme results in pure exploitation and may lead to misbehavior due to noise in stochastic environments. Thus, it is critical to balancing exploration and exploitation in value function estimation. The Boltzmann softmax operator has a greater capability in exploring potential action-values. However, it does not satisfy the non-expansion property, and its direct use may fail to converge even in value iteration. In this paper, we propose to update the value function with dynamic Boltzmann softmax (DBS) operator in value function estimation, which has good convergence property in the setting of planning and learning. Moreover, we prove that dynamic Boltzmann softmax updates can eliminate the overestimation phenomenon introduced by the hard max operator. Experimental results on GridWorld show that the DBS operator enables convergence and a better trade-off between exploration and exploitation in value function estimation. Finally, we propose the DBS-DQN algorithm by generalizing the dynamic Boltzmann softmax update in deep Q-network, which outperforms DQN substantially in 40 out of 49 Atari games.

  Click for Model/Code and Paper