We propose a novel modification of the standard upper confidence bound (UCB) method for the stochastic multi-armed bandit (MAB) problem which tunes the confidence bound of a given bandit based on its distance to others. Our UCB distance tuning (UCB-DT) formulation enables improved performance as measured by expected regret by preventing the MAB algorithm from focusing on non-optimal bandits which is a well-known deficiency of standard UCB. "Distance tuning" of the standard UCB is done using a proposed distance measure, which we call bandit distance, that is parameterizable and which therefore can be optimized to control the transition rate from exploration to exploitation based on problem requirements. We empirically demonstrate increased performance of UCB-DT versus many existing state-of-the-art methods which use the UCB formulation for the MAB problem. Our contribution also includes the development of a conceptual tool called the "Exploration Bargain Point" which gives insights into the tradeoffs between exploration and exploitation. We argue that the Exploration Bargain Point provides an intuitive perspective that is useful for comparatively analyzing the performance of UCB-based methods.
相關內容
The growth of deep reinforcement learning (RL) has brought multiple exciting tools and methods to the field. This rapid expansion makes it important to understand the interplay between individual elements of the RL toolbox. We approach this task from an empirical perspective by conducting a study in the continuous control setting. We present multiple insights of fundamental nature, including: an average of multiple actors trained from the same data boosts performance; the existing methods are unstable across training runs, epochs of training, and evaluation runs; a commonly used additive action noise is not required for effective training; a strategy based on posterior sampling explores better than the approximated UCB combined with the weighted Bellman backup; the weighted Bellman backup alone cannot replace the clipped double Q-Learning; the critics' initialization plays the major role in ensemble-based actor-critic exploration. As a conclusion, we show how existing tools can be brought together in a novel way, giving rise to the Ensemble Deep Deterministic Policy Gradients (ED2) method, to yield state-of-the-art results on continuous control tasks from OpenAI Gym MuJoCo. From the practical side, ED2 is conceptually straightforward, easy to code, and does not require knowledge outside of the existing RL toolbox.
We consider pessimistic bilevel stochastic programs in which the follower maximizes over a fixed compact convex set a strictly convex quadratic function, whose Hessian depends on the leader's decision. The resulting random variable is evaluated by a convex risk measure. Under assumptions including real analyticity of the lower-level goal function, we prove existence of optimal solutions. We discuss an alternate model where the leader hedges against optimal lower-level solutions, and show that in this case solvability can be guaranteed under weaker conditions both in a deterministic and in a stochastic setting. The approach is applied to a mechanical shape optimization problem in which the leader decides on an optimal material distribution to minimize a tracking-type cost functional, whereas the follower chooses forces from an admissible set to maximize a compliance objective. The material distribution is considered to be stochastically perturbed in the actual construction phase. Computational results illustrate the bilevel optimization concept and demonstrate the interplay of follower and leader in shape design and testing.
Contextual multi-armed bandits are classical models in reinforcement learning for sequential decision-making associated with individual information. A widely-used policy for bandits is Thompson Sampling, where samples from a data-driven probabilistic belief about unknown parameters are used to select the control actions. For this computationally fast algorithm, performance analyses are available under full context-observations. However, little is known for problems that contexts are not fully observed. We propose a Thompson Sampling algorithm for partially observable contextual multi-armed bandits, and establish theoretical performance guarantees. Technically, we show that the regret of the presented policy scales logarithmically with time and the number of arms, and linearly with the dimension. Further, we establish rates of learning unknown parameters, and provide illustrative numerical analyses.
Offline policy learning (OPL) leverages existing data collected a priori for policy optimization without any active exploration. Despite the prevalence and recent interest in this problem, its theoretical and algorithmic foundations in function approximation settings remain under-developed. In this paper, we consider this problem on the axes of distributional shift, optimization, and generalization in offline contextual bandits with neural networks. In particular, we propose a provably efficient offline contextual bandit with neural network function approximation that does not require any functional assumption on the reward. We show that our method provably generalizes over unseen contexts under a milder condition for distributional shift than the existing OPL works. Notably, unlike any other OPL method, our method learns from the offline data in an online manner using stochastic gradient descent, allowing us to leverage the benefits of online learning into an offline setting. Moreover, we show that our method is more computationally efficient and has a better dependence on the effective dimension of the neural network than an online counterpart. Finally, we demonstrate the empirical effectiveness of our method in a range of synthetic and real-world OPL problems.
In blind compression of quantum states, a sender Alice is given a specimen of a quantum state $\rho$ drawn from a known ensemble (but without knowing what $\rho$ is), and she transmits sufficient quantum data to a receiver Bob so that he can decode a near perfect specimen of $\rho$. For many such states drawn iid from the ensemble, the asymptotically achievable rate is the number of qubits required to be transmitted per state. The Holevo information is a lower bound for the achievable rate, and is attained for pure state ensembles, or in the related scenario of entanglement-assisted visible compression of mixed states wherein Alice knows what state is drawn. In this paper, we prove a general and robust lower bound on the achievable rate for ensembles of classical states, which holds even in the least demanding setting when Alice and Bob share free entanglement and a constant per-copy error is allowed. We apply the bound to a specific ensemble of only two states and prove a near-maximal separation (saturating the dimension bound in leading order) between the best achievable rate and the Holevo information for constant error. This also implies that the ensemble is incompressible -- compression does not reduce the communication cost by much. Since the states are classical, the observed incompressibility is not fundamentally quantum mechanical. We lower bound the difference between the achievable rate and the Holevo information in terms of quantitative limitations to clone the specimen or to distinguish the two classical states.
The fidelity bandits problem is a variant of the $K$-armed bandit problem in which the reward of each arm is augmented by a fidelity reward that provides the player with an additional payoff depending on how 'loyal' the player has been to that arm in the past. We propose two models for fidelity. In the loyalty-points model the amount of extra reward depends on the number of times the arm has previously been played. In the subscription model the additional reward depends on the current number of consecutive draws of the arm. We consider both stochastic and adversarial problems. Since single-arm strategies are not always optimal in stochastic problems, the notion of regret in the adversarial setting needs careful adjustment. We introduce three possible notions of regret and investigate which can be bounded sublinearly. We study in detail the special cases of increasing, decreasing and coupon (where the player gets an additional reward after every $m$ plays of an arm) fidelity rewards. For the models which do not necessarily enjoy sublinear regret, we provide a worst case lower bound. For those models which exhibit sublinear regret, we provide algorithms and bound their regret.
We establish generalization error bounds for stochastic gradient Langevin dynamics (SGLD) with constant learning rate under the assumptions of dissipativity and smoothness, a setting that has received increased attention in the sampling/optimization literature. Unlike existing bounds for SGLD in non-convex settings, ours are time-independent and decay to zero as the sample size increases. Using the framework of uniform stability, we establish time-independent bounds by exploiting the Wasserstein contraction property of the Langevin diffusion, which also allows us to circumvent the need to bound gradients using Lipschitz-like assumptions. Our analysis also supports variants of SGLD that use different discretization methods, incorporate Euclidean projections, or use non-isotropic noise.
The difficulty in specifying rewards for many real-world problems has led to an increased focus on learning rewards from human feedback, such as demonstrations. However, there are often many different reward functions that explain the human feedback, leaving agents with uncertainty over what the true reward function is. While most policy optimization approaches handle this uncertainty by optimizing for expected performance, many applications demand risk-averse behavior. We derive a novel policy gradient-style robust optimization approach, PG-BROIL, that optimizes a soft-robust objective that balances expected performance and risk. To the best of our knowledge, PG-BROIL is the first policy optimization algorithm robust to a distribution of reward hypotheses which can scale to continuous MDPs. Results suggest that PG-BROIL can produce a family of behaviors ranging from risk-neutral to risk-averse and outperforms state-of-the-art imitation learning algorithms when learning from ambiguous demonstrations by hedging against uncertainty, rather than seeking to uniquely identify the demonstrator's reward function.
Efficient exploration remains a major challenge for reinforcement learning. One reason is that the variability of the returns often depends on the current state and action, and is therefore heteroscedastic. Classical exploration strategies such as upper confidence bound algorithms and Thompson sampling fail to appropriately account for heteroscedasticity, even in the bandit setting. Motivated by recent findings that address this issue in bandits, we propose to use Information-Directed Sampling (IDS) for exploration in reinforcement learning. As our main contribution, we build on recent advances in distributional reinforcement learning and propose a novel, tractable approximation of IDS for deep Q-learning. The resulting exploration strategy explicitly accounts for both parametric uncertainty and heteroscedastic observation noise. We evaluate our method on Atari games and demonstrate a significant improvement over alternative approaches.
We consider the exploration-exploitation trade-off in reinforcement learning and we show that an agent imbued with a risk-seeking utility function is able to explore efficiently, as measured by regret. The parameter that controls how risk-seeking the agent is can be optimized exactly, or annealed according to a schedule. We call the resulting algorithm K-learning and show that the corresponding K-values are optimistic for the expected Q-values at each state-action pair. The K-values induce a natural Boltzmann exploration policy for which the `temperature' parameter is equal to the risk-seeking parameter. This policy achieves an expected regret bound of $\tilde O(L^{3/2} \sqrt{S A T})$, where $L$ is the time horizon, $S$ is the number of states, $A$ is the number of actions, and $T$ is the total number of elapsed time-steps. This bound is only a factor of $L$ larger than the established lower bound. K-learning can be interpreted as mirror descent in the policy space, and it is similar to other well-known methods in the literature, including Q-learning, soft-Q-learning, and maximum entropy policy gradient, and is closely related to optimism and count based exploration methods. K-learning is simple to implement, as it only requires adding a bonus to the reward at each state-action and then solving a Bellman equation. We conclude with a numerical example demonstrating that K-learning is competitive with other state-of-the-art algorithms in practice.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191