Generating interpretable reinforcement learning policies using genetic programming

Interpretable policies for reinforcement learning by genetic programming

Introduction Reinforcement learning formalizes the concept of learning from interactions.1 Broadly, reinforcement learning focuses on a setting in which an agent (decision maker) sequentially interacts with an environment that is partially unknown to them. At each stage, the agent takes an action and receives a reward. The objective of the agent is to maximize rewards accumulated in the long run. There are many situations in health care where decisions are made sequentially for which reinforcement learning approaches could prove useful for decision making. Throughout this article, we consider treatment prescription as an archetypical example to connect reinforcement learning concepts to a health care setting. In this setting, the care provider, the prescribed treatment, and the patients can be viewed as the agent, the action, and the environment, respectively, as depicted in Figure 1.Figure 1: Sequential treatment of AKI or CKD complications modeled as a reinforcement learning problem.Background In this section, with the objective of making reinforcement learning literature more accessible to a clinical audience, we briefly introduce related fundamental concepts and approaches. We refer the interested reader to Sutton and Barto1 for a comprehensive introduction to reinforcement learning. Markov Decision Processes Markov decision processes (MDPs) are a formalism of the sequential decision-making problem that has been central to the theoretical and practical advancements of reinforcement learning. In each stage of an MDP, the agent observes the state of the environment and takes an action, which, in turn, results in a change of the state. This change of state is assumed to be probabilistic with the next state being determined only by the preceding state, the chosen action, and the transition probability. The agent also receives a reward that is a function of the taken action, the preceding state, and the subsequent state. In an MDP, the objective of the agent is to maximize the return defined as the reward accumulated over a time horizon. In some applications, it is common to consider the horizon to be infinite, in which case the future rewards are discounted by a factor smaller than one. The selection of action by the agent on the basis of the observed state is known as the policy. More formally, a policy is a probabilistic mapping from states to each possible action. Because the policy and the reward are a function of the state, it is critical to estimate the utility of being in a certain state. More specifically, the value function is defined as the expected return starting from a given state under the chosen policy. Under this formalism, the objective of the agent is to find the optimal policy that maximizes the value function for all states. Reinforcement Learning Methods Action-value methods are a class of reinforcement learning methods in which the actions are chosen on the basis of the estimation of their long-term value. A prominent example of an action-value method is Q-learning in which the agent iteratively takes actions with the highest estimated values and updates the action-state value function on the basis of new observations. Policy gradient methods are another class of reinforcement learning methods that seek to optimize the policy directly instead of choosing actions on the basis of their respective estimated value. Such methods could be advantageous in health care applications that entail a large number of possible actions, e.g., when recommending a wide range of drug dosages or treatment options. Clinical Applications Reinforcement learning frameworks and methods are broadly applicable to clinical settings in which decisions are made sequentially. A prominent clinical application of reinforcement learning is for treatment recommendation, which has been studied across a variety of diseases and treatments including radiation and chemotherapy for cancer, brain stimulation for epilepsy, and treatment strategies for sepsis.2–5 In such treatment recommendation settings, a policy is commonly known as a dynamic treatment regime. There are various other clinical applications of reinforcement learning including diagnosis, medical imaging, and decision support tools (see refs. 2–5 and the references therein). Reinforcement Learning in Nephrology Although there have been recent applications of machine learning in nephrology,6,7 to the best of the authors' knowledge, the application of reinforcement learning to nephrology has been primarily limited to optimizing the erythropoietin dosage in hemodialysis patients.8,9 However, there are other settings where reinforcement learning has the potential to improve patient care in nephrology. For example, reinforcement learning methods can be adopted in the treatment of the complications of AKI or CKD (Figure 1). In this problem, the state models the conditions of the patient (e.g., vital signs, laboratory test results including urine and blood tests, and urine output measurements). The action refers to the treatment options (e.g., the dosage of medications such as sodium polystyrene sulfonate, and hemodialysis). The reward models the improvement in patient conditions. Similarly, reinforcement learning can help automate and optimize the dosage of immunosuppressive drugs in kidney transplants. Challenges and Opportunities Despite the success of reinforcement learning in several simplified clinical settings, their large-scale application to patient care faces several open challenges. The complexity of human biology complicates modeling clinical decision making as a reinforcement learning problem. The state space in such settings is often enormous, which could make a purely computational approach infeasible. Moreover, modeling all potential objectives a priori as a reward function may not be feasible. To overcome these challenges and realize the potential of reinforcement learning, clinical insight can play a pivotal role. More specifically, restricting the state space to only include highly relevant clinical variables could greatly reduce the computational complexity. Furthermore, using inverse reinforcement learning,2 relevant reward functions can be learned from retrospective studies assuming the optimality of clinical decisions. Another critical challenge is addressing moral and ethical concerns. It is imperative to ensure that reinforcement learning methods do not cause harm to the patient. To this end, there exists a need for a thorough validation of such methods before their use in patient care. Hence, there is a need to go beyond retrospective studies that have been used for the proof of concept of most existing reinforcement learning methods in health care applications.2,3 The lessons learned from the success of reinforcement learning in other application areas (e.g., self-driving cars) can help navigate the path to realizing its potential in health care. Accessible open-source simulation environments that enable researchers to compare various approaches are essential to the field of reinforcement learning. OpenAI Gym is currently the leading toolkit containing a wide range of simulated environments, e.g., surgical robotics.10 The development of high-quality and reliable simulation environments for nephrology and other health care applications can facilitate the development and validation of reinforcement learning methods beyond limited retrospective studies. The adoption of methods validated in such simulation environments in actual clinical settings will require clinicians' oversight. Similar to how self-driving cars require a human driver to ensure collision avoidance, clinicians' oversight is critical to ensure the safety of the patients, especially in the early stages of the adoption of reinforcement learning methods. The data from clinicians' decisions (e.g., overruling the automated treatment recommendation) can be used to improve the reliability of autonomous systems over time and reduce the burden of clinicians' oversight.

Active screening is a common approach in controlling the spread of recurring infectious diseases such as tuberculosis and influenza. In this approach, health workers periodically select a subset of population for screening. However, given the limited number of health workers, only a small subset of the population can be visited in any given time period. Given the recurrent nature of the disease and rapid spreading, the goal is to minimize the number of infections over a long time horizon. Active screening can be formalized as a sequential combinatorial optimization over the network of people and their connections. The main computational challenges in this formalization arise from i) the combinatorial nature of the problem, ii) the need of sequential planning and iii) the uncertainties in the infectiousness states of the population. Previous works on active screening fail to scale to large time horizon while fully considering the future effect of current interventions. In this paper, we propose a novel reinforcement learning (RL) approach based on Deep Q-Networks (DQN), with several innovative adaptations that are designed to address the above challenges. First, we use graph convolutional networks (GCNs) to represent the Q-function that exploit the node correlations of the underlying contact network. Second, to avoid solving a combinatorial optimization problem in each time period, we decompose the node set selection as a sub-sequence of decisions, and further design a two-level RL framework that solves the problem in a hierarchical way. Finally, to speed-up the slow convergence of RL which arises from reward sparseness, we incorporate ideas from curriculum learning into our hierarchical RL approach. We evaluate our RL algorithm on several real-world networks. Results show that our RL algorithm can scale up to 10 times the problem size of state-of-the-art (the variant that considers the effect of future interventions but unscalable) in terms of planning time horizon. Meanwhile, it outperforms state-of-the-art (the variant that scales up but does not consider the effect of future interventions) by up to 33% in solution quality.

Data Mining is the process of extracting patterns from data. Two general avenues of research in the intersecting areas of agents and data mining can be distinguished. The first approach is concerned with mining an agent's observation data in order to extract patterns, categorize environment states, and/or make predictions of future states. In this setting, data is normally available as a batch, and the agent's actions and goals are often independent of the data mining task. The data collection is mainly considered as a side effect of the agent's activities. Machine learning techniques applied in such situations fall into the class of supervised learning. In contrast, the second scenario occurs where an agent is actively performing the data mining, and is responsible for the data collection itself. For example, a mobile network agent is acquiring and processing data (where the acquisition may incur a certain cost), or a mobile sensor agent is moving in a (perhaps hostile) environment, collecting and processing sensor readings. In these settings, the tasks of the agent and the data mining are highly intertwined and interdependent (or even identical). Supervised learning is not a suitable technique for these cases. Reinforcement Learning (RL) enables an agent to learn from experience (in form of reward and punishment for explorative actions) and adapt to new situations, without a teacher. RL is an ideal learning technique for these data mining scenarios, because it fits the agent paradigm of continuous sensing and acting, and the RL agent is able to learn to make decisions on the sampling of the environment which provides the data. Nevertheless, RL still suffers from scalability problems, which have prevented its successful use in many complex real-world domains. The more complex the tasks, the longer it takes a reinforcement learning algorithm to converge to a good solution. For many real-world tasks, human expert knowledge is available. For example, human experts have developed heuristics that help them in planning and scheduling resources in their work place. However, this domain knowledge is often rough and incomplete. When the domain knowledge is used directly by an automated expert system, the solutions are often sub-optimal, due to the incompleteness of the knowledge, the uncertainty of environments, and the possibility to encounter unexpected situations. RL, on the other hand, can overcome the weaknesses of the heuristic domain knowledge and produce optimal solutions. In the talk we propose two techniques, which represent first steps in the area of knowledge-based RL (KBRL). The first technique [1] uses high-level STRIPS operator knowledge in reward shaping to focus the search for the optimal policy. Empirical results show that the plan-based reward shaping approach outperforms other RL techniques, including alternative manual and MDP-based reward shaping when it is used in its basic form. We showed that MDP-based reward shaping may fail and successful experiments with STRIPS-based shaping suggest modifications which can overcome encountered problems. The STRIPSbased method we propose allows expressing the same domain knowledge in a different way and the domain expert can choose whether to define an MDP or STRIPS planning task. We also evaluated the robustness of the proposed STRIPS-based technique to errors in the plan knowledge. In case that STRIPS knowledge is not available, we propose a second technique [2] that shapes the reward with hierarchical tile coding. Where the Q-function is represented with low-level tile coding, a V-function with coarser tile coding can be learned in parallel and used to approximate the potential for ground states. In the context of data mining, our KBRL approaches can also be used for any data collection task where the acquisition of data may incur considerable cost. In addition, observing the data collection agent in specific scenarios may lead to new insights into optimal data collection behaviour in the respective domains. In future work, we intend to demonstrate and evaluate our techniques on concrete real-world data mining applications.

To find the path that minimizes the time to navigate between two given points in a fluid flow is known as Zermelo's problem. Here, we investigate it by using a Reinforcement Learning (RL) approach for the case of a vessel that has a slip velocity with fixed intensity, Vs, but variable direction and navigating in a 2D turbulent sea. We show that an Actor-Critic RL algorithm is able to find quasioptimal solutions for both time-independent and chaotically evolving flow configurations. For the frozen case, we also compared the results with strategies obtained analytically from continuous Optimal Navigation (ON) protocols. We show that for our application, ON solutions are unstable for the typical duration of the navigation process and are, therefore, not useful in practice. On the other hand, RL solutions are much more robust with respect to small changes in the initial conditions and to external noise, even when Vs is much smaller than the maximum flow velocity. Furthermore, we show how the RL approach is able to take advantage of the flow properties in order to reach the target, especially when the steering speed is small.

Reinforcement learning for a class of continuous-time input constrained optimal control problems

Aim and ApproachLarge cooling water system has great potential of energy saving due to improper operation control. The coupled hydraulic and thermodynamic characteristics among chillers, pumps, and cooling towers lead to the difficulty of cooling water system operation. This study investigates an advanced control approach of multi-agent reinforcement learning (RL) for a large cooling water system. The multiple RL agents interact with each other to deal with the complex coupled characteristics in the cooling water system. The proposed approach is implemented in a real-world cooling water system and validated with experiments.Scientific Innovation and RelevanceIn this study, a multi-agent RL approach is proposed to separately control the chillers, pumps, and cooling towers in the cooling water system. The soft actor critic algorithm is used in the RL agent, as it is an efficient and advanced RL algorithm for continuous and discrete control problems. The agents are designed to have private states, private actions, and shared rewards. The reward functions are carefully designed to balance the agents for global optima.To provide a realistic environment for RL training, a detailed physical modeling platform of a cooling water system is established and validated. This platform is capable of hydraulic and thermodynamic calculation of cooling water systems with customized numbers and characteristics of chillers, pumps, and cooling towers, which is scalable for different cooling water systems.The proposed multi-agent RL controller is implemented in the building automation system of a real cooling water system. A one-month experimental study of the multi-agent RL performance is conducted. The proposed multi-agent RL controller is compared with a single-agent RL controller and a rule-based controller through real-world experiments to demonstrate the energy performance. This is a pilot study of multi-agent RL approach to real-world cooling water system control.Preliminary Results and ConclusionsThe preliminary simulation study is conducted in the established cooling water system model. The electricity consumption simulation error is 4.0% in coefficient of variation of mean absolute error (CVMAE) compared with the measured 15-minute electricity consumption. The 6-month simulation shows that the single-agent RL can save 6.2% energy from the rule-based control, while the multi-agent RL can save 7.2% electricity consumption of the cooling water system. In the following study, the multi-agent RL implementation will be practiced with more experiments and applications for cooling water system control.

In the ever-changing world of decision-making, when game theory and reinforcement learning(RL) come together, they create a fascinating combination that shows a new way to solve complex problems in many different fields. The combination of game theory and RL is a powerful convergence that opens up a hopeful new frontier for dealing with difficult decision-making problems in many different fields. Research on the convergence of game theory and RL has shown to be beneficial, providing essential insights into challenging decision-making issues in various disciplines. This study investigates the recent developments of game theory and RL approaches through a systematic review and highlights the significance of game theory in boosting reinforcement algorithms and increasing the interaction of autonomous vehicles, safeguarding edge caching, and more. It offers a thorough account of the developments at the confluence of game theory and RL. The reviewed papers mainly focus on broad themes and address three important research questions: the impact of game theory on multi-agent reinforcement learning (MARL), the significant contributions of game theory to RL, and the significant impact areas. Following the methodology, search outcomes, and study areas is a discussion on game theory-related terminology, followed by study findings. The review’s conclusions are offered with ideas for further study and open research questions. The importance of game theory in advancing MARL, the potential of game theory in promoting RL strategies, and the opportunities for combining game theory and RL in cutting-edge fields like mobile edge caching and cyber-physical systems(CPS) are all emphasized in the conclusion. This review article advances our knowledge of the theoretical underpinnings and real-world applications of game theory and RL, laying the groundwork for future improvements in decision-making techniques and algorithms.

Model-based contextual policy search for data-efficient generalization of robot skills

Experimental research such as cell cultures and carbon nanotube (CNT) growth are largely governed by following predefined execution protocols with fine-tuned control of parameters. There are promising opportunities to apply reinforcement learning (RL), an established learning technique in the area of artificial intelligence, in order to automate CNT growth process and accelerate related scientific breakthroughs in material discovery. Although there are benefits in RL-based exploration and exploitation methodologies, there are also challenges in developing relevant learning policies in experimental settings relating to CNT growth. In this paper, we present a novel data-driven RL approach for assisting experimental CNT growth. Our approach focuses on developing an RL model to learn from simulation-based images and characteristics of temporal CNT growth, considering various growth parameters. Our RL model learns from CNT growth variation in a simulation-based environment where critical control parameters, such as density, growth rate, tube radius, tube stiffness, and Van der Waals forces, are used. Our RL model enables CNT growth automation in order to explore of a wider range of growth conditions, and improve reproducibility. The ultimate goal of our RL model is to achieve desired CNT growth by dynamically controlling growth parameters throughout a sequence of experiments. We evaluate the effectiveness of our RL approach by measuring the improvement in the maximum compressive strength of a carbon nanotube ‘with’ and ‘without’ the RL model. Our results show the effectiveness of course correction recommended by our RL approach when controlling the parameters of angular deviation and rate of growth of carbon nanotubes, when compared against non-regulated CNT growth.

Learning policies that generalize across multiple tasks is an important and challenging research topic in reinforcement learning and robotics. Training individual policies for every single potential task is often impractical, especially for continuous task variations, requiring more principled approaches to share and transfer knowledge among similar tasks. We present a novel approach for learning a nonlinear feedback policy that generalizes across multiple tasks. The key idea is to define a parametrized policy as a function of both the state and the task, which allows learning a single policy that generalizes across multiple known and unknown tasks. Applications of our novel approach to reinforcement and imitation learning in real-robot experiments are shown.

A reinforcement learning and predictive analytics approach for enhancing credit assessment in manufacturing

Mobile Edge Computing (MEC) is a promising paradigm to support high-quality time-sensitive applications. In this paper, we investigate the service migration (i.e., whether, when, and where to migrate the services) to seamlessly serve mobile users in small-cell MEC systems. The service migration is formulated as an optimization problem to minimize the long-term system average delay that consists of queuing, communication, and migration delays. Considering the dynamic user mobility and network conditions, the formulated problem is non-convex and difficult to solve in real time. To this end, we propose a Mobility-aware Service Migration scheme, named MSM, to make real-time decisions on service migrations by utilizing reinforcement learning (RL) approaches. Specifically, we first design a user classification mechanism based on users’ mobility patterns to reduce the complexity of decision-making. We then formulate the service migration as a Markov decision process and devise an RL-based framework to make service migration decisions in real time in the dynamic MEC environment. Extensive data-driven experiments demonstrate the efficacy of MSM in reducing the system average delay.

The emergence of various applications is driving the continuous development of 5G mobile networks as infrastructure. To better support real-time wireless services, the low-latency communication is the current research focus. However, the existing radio resource scheduling methods cannot guarantee the strict delay constraint of the low-latency communication. Therefore, a reinforcement learning (RL) approach of radio spectrum resource scheduling strategy is proposed, which can guarantee the low-latency constraint when spectrum resources are insufficient. The Q-learning algorithm is used to approximate the optimal goal of RL. To speed up the learning, the deep neural network (DNN) is used to train the learning parameters. Simulation results show that the strategy converges quickly and has satisfactory results for mobile networks with a high load of spectrum resources.

Dialogue policy plays an important role in task-oriented spoken dialogue systems. It determines how to respond to users. The recently proposed deep reinforcement learning (DRL) approaches have been used for policy optimization. However, these deep models are still challenging for two reasons: first, many DRL-based policies are not sample efficient; and second, most models do not have the capability of policy transfer between different domains. In this paper, we propose a universal framework, AgentGraph, to tackle these two problems. The proposed AgentGraph is the combination of graph neural network (GNN) based architecture and DRL-based algorithm. It can be regarded as one of the multi-agent reinforcement learning approaches. Each agent corresponds to a node in a graph, which is defined according to the dialogue domain ontology. When making a decision, each agent can communicate with its neighbors on the graph. Under AgentGraph framework, we further propose dual GNN-based dialogue policy, which implicitly decomposes the decision in each turn into a high-level global decision and a low-level local decision. Experiments show that AgentGraph models significantly outperform traditional reinforcement learning approaches on most of the 18 tasks of the PyDial benchmark. Moreover, when transferred from the source task to a target task, these models not only have acceptable initial performance but also converge much faster on the target task.