Abstract
Reinforcement learning based on the deep neural network has attracted much attention and has been widely used in realworld applications. However, the blackbox property limits its usage from applying in highstake areas, such as manufacture and healthcare. To deal with this problem, some researchers resort to the interpretable control policy generation algorithm. The basic idea is to use an interpretable model, such as treebased genetic programming, to extract policy from other black box modes, such as neural networks. Following this idea, in this paper, we try yet another form of the genetic programming technique, evolutionary feature synthesis, to extract control policy from the neural network. We also propose an evolutionary method to optimize the operator set of the control policy for each specific problem automatically. Moreover, a policy simplification strategy is also introduced. We conduct experiments on four reinforcement learning environments. The experiment results reveal that evolutionary feature synthesis can achieve better performance than treebased genetic programming to extract policy from the neural network with comparable interpretability.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Introduction
Reinforcement learning [31] has shown its extraordinary performance in computer games [22] and other realworld applications [29]. The neural network is widely used as a dominant model to solve reinforcement learning problems. Generally, we call these methods deep reinforcement learning algorithms, since these algorithms use a deep neural network as the value function approximator or the policy function approximator. Deep qlearning (DQN) [22], double DQN [9], dueling DQN (DDQN) [36] are prestigious algorithms that train a deep neural network for reinforcement learning problems. However, the blackbox property of the deep neural network prevents DNN to be directly used in the highstake scenarios [33]. Therefore, building an interpretable model is essential, and even more priority than interpreting the blackbox model in the current machine learning field [27].
There are a variety of ways to build interpretable models [15, 34]. Among them, the genetic programming (GP), which builds a symbolic expression as an explainable model through the genetic algorithm, is a promising one. Recently, GP has been applied to reinforcement learning. The idea is to evolve an explainable model to extract the policy from the deep neural network. In [11], an explainable reinforcement learning policy model is built by using the treebased genetic programming (GP) [24] algorithm. However, it is argued in [11] that it is hard for GP to mimic the behavior of the deep neural network. Therefore, in that paper, DNN is used as a surrogate model, and then GP is used to evolving strategy on that model. This method is far from meeting the requirements of what we desired, deriving a policy from a pretrained neural network model built by the cuttingedge algorithms such as DQN [22] or ActorCritic [21] algorithms.
In this paper, we propose a method to extract policy from the pretrained deep neural network based on the evolution feature synthesis (EFS) algorithm [3]. Our method first generates a behavior sequence by running the neural network in the real environment and then evolve a set of regressors to mimic that behavior sequence as similar as possible. We improve the performance using an evolutionary method to automatically select operators from a large set of predefined operators without domain experts involved. Moreover, we introduce a policy simplification approach to balance the tradeoff between interpretability and performance. We conduct comparative experiments on the CartPole [31], the Acrobot [30], the MountainCar [23], and the Industrial Benchmark [10] environments. All these experimental results prove that the policy extracted by EFS outperforms the policy extracted by GP with similar interpretability. Furthermore, the operator optimization strategy has significantly improved final rewards and eventually makes EFS surpass the neural network and common interpretable machine learning algorithms in all environments.
The main contributions of our work are summarized as follows:

We derive the latent policy from the pretrained deep neural network by EFS. Our experimental results demonstrate that EFS can achieve comparable performance to the deep neural network.

We design an operator optimization strategy based on the evolutionary algorithm. This strategy can automatically select appropriate operators from a large set of predefined operators without additional domain knowledge.

We propose a progressive simplification method to construct a series of control policies based on the best EFS policy model to balance the tradeoff between the model complexity and performance.

We evaluate our algorithm on the CartPole, the Acrobot, the MountainCar, and the Industrial Benchmark problems. The experimental results are compared with the results of GP, DNN, Linear regression, Decision tree, KNearest neighbors (KNN). The experimental results show that our algorithm achieves better performance than GP, DNN, and common interpretable machine learning algorithms with comparable interpretability.
The rest of this paper is organized as follows: the next section introduces the related works of reinforcement learning based on genetic programming algorithms. The following section gives a brief introduction to EFS. The next section presents the details of the EFSbased policy derivation algorithm. The following section presents experimental settings and experimental results of the proposed algorithm. Finally, we conclude this paper with some future works in the last section.
Related work
Reinforcement learning by genetic programming can retrospect to [17], which uses GP to solve the broom balancing problem. Followed by [28], an equation is generated by GP to solve the movable inverted pendulum problem. Subsequently, genetic network programming (GNP) [14] is proposed to surpass GP on the foodcollecting problem [12]. Soon after, the performance of GNP is enhanced by QLearning [20] and by utilizing the information obtained during tasks [19].
Recently, the value function approximation by GP becomes a hot spot in the evolutionary reinforcement learning domain. Reference [25] introduced a method to get a nearoptimal value function on an illustrative Markov decision process (MDP) environment. Lately, [18] proposed an approach to calculate the symbolic value function in a policy iteration way. GP based value function approximation approach has been proven that it can obtain an accurate value function even for a small batch of data [6, 7].
The incorporation of the modelbased reinforcement learning and the symbolic regression algorithm has been studied in [11]. Reference [11] introduced a method of evolving interpretable policies based on GP. A neural network that builds based on trajectory data is used as a world model to evaluate the fitness value of generated symbolic policies.
In addition to leverage genetic programming to generate interpretable policies, the influence of the smooth property of symbolic approximation function in reinforcement learning has also been studied. References [1] and [2] proposed a symbolic regression method to obtain a smooth vfunction from the original value function, and experiment results reveal that policy derived from the smooth value function performs much better than the policy derived from the original value function.
However, to the best of our knowledge, up to now, GP is hard to derive policy from the deep neural network agent [11]. EFS is the first model that exhibits that the genetic programming algorithm can derive policy directly from DNN to achieve comparable performance with stronger interpretability.
Evolutionary feature synthesis
EFS [3] is a novel genetic programming algorithm, which evolves nonlinear features to compose a linear model iteratively. The search space of EFS is all possible nonlinear features which are composed of variables and operators in a predefined set. The goal of the evolution process is to find a set of nonlinear features, which constitute a linear model that can perform well on a specific problem. After a feature initialization stage, the evolution process of EFS comprises three stages, i.e., the feature composition stage, the feature importance evaluation stage, and the feature selection stage.
First, in the initialization stage, each original variable x in the training data X is added to the feature set U. From the perspective the evolutionary computation, the feature set U is also called as a population of features. The original variables are marked by a distinctive mark so that they will not be removed from the population in the following process.
In the feature composition stage, an operator o is randomly selected from the operator set O. If o is a unary operator, such as log or square, then one feature u is selected from the population U and then the operator o is used to compose a new feature o(u). If o is a binary operator, such as \(+\) or \(*\), then two features \(u_1,u_2\) are selected from the population U, and then the operator o is used to compose a new feature \(o(u_1,u_2)\). Suppose the population size is P, then \(PX\) features should be generated in this stage because original variables remain unchanged.
In the stage of the feature importance evaluation, the importance of each feature is calculated by a score function g. A typical score function is the fregression function
where \({\bar{x}}\) represents the mean value of feature values x, \({\bar{y}}\) represents the mean value of target values y. \(\sigma _x\) represents the variance of x, and \(\sigma _y\) represents the variance of y. Y represents the count of data items.
In the stage of feature selection, features are ranked by the importance value. Then the top P features are selected as the basis of the linear model, and other features are discarded.
Finally, in the stage of model construction, a linear model is composed of multiple nonlinear features P. The parameters of the linear model are determined by the least square method. The overall process is presented in Fig. 1 and Algorithm 1. More details about Algorithm 1 are referred to [3].
EFSbased policy derivation
In this section, we propose a new policy derivation method based on EFS, extracting the wellperformed policy from the deep neural network by evolutionarily constructing a linear model with nonlinear features. Formally, there are two essential components of our EFS based reinforcement learning algorithm.

Deep neural network: DNN is used to approximate the optimal value function [22] or the policy function [32] of an environment. Through DNN, we can generate behavior sequences S, which comprise states s and actions a by simulating in the real environment. Then, EFS can utilize these behavior sequences \((s,a) \in S\) to generate an interpretable policy model.

EFS model set: EFS model set is composed of several EFS models \(m \in M\). Each EFS model is to mimic a part of the behavior sequences S to decide which action a should be chosen at a specific state s. The size of the model set M is equal to the size of the action space A. The details of the training and the prediction process are described in the following section.
The policy derivation process is composed of five subprocesses.

Data preparation: In the data preparation process, we use a pretrained neural network to generate optimal decision schemes in the reinforcement learning environment under different circumstances.

Operator optimization: Operator optimization is an optional process, which preselect an optimal subset of operators by the genetic algorithm based on smallscale experiments.

EFS evolution: In the EFS evolution process, we evolve a set of EFS regressors based on the training data obtained in the preparation stage.

EFS ensembling: After getting a set of regressors, we aggregate these regressors as a single model based on a simple decision function.

Policy simplification: Policy simplification is also an optional operation, which can balance the tradeoff between complexity and accuracy.
In the following sections, we first present the essential process of our algorithm, and then we introduce two optional strategies to achieve a better result. The workflow of our algorithm is presented in Fig. 2 and Algorithm 2.
Data preparation
In the behavior sequences preparation stage, the most convenient way is to leverage the pretrained neural network model to generate behavior sequences in the real environment. First, it is necessary to emphasize that all of environments following the Markov property, i.e. \(s,r \leftarrow \text {environment}(s,a)\). Then, we can use the neural network to generate an action a at each state s
where \(\text {DNN}\) represents the neural network. In QLearning, \(\text {DNN}(s,a)\) generates a vector that represents the estimated reward when performing different actions a at the current state s. The action corresponding to the best estimate reward is selected as the real action. Each stateaction pair is stored in the behavior sequences S, and the behavior sequences generation should be continuously executed until the predefined e episode number is met. Figure 3 presents a schematic of the data preparation process based on the neural network. In Algorithm 2, lines 1–12 presents the details of the data preparation process.
EFS evolution
In the EFS evolution process, the target is to evolve a set of regressors M to mimic the behavior sequences as well as possible. In other words, the evolution target is to minimize the loss value of each EFS regressor \(f_a \in M\) on the behavior sequences
where \(f_a(s_i)\) is the predicted value of the EFS regressor \(f_a\). \(I_a(s_i)\) is an indicator function which indicates whether the action a has been chosen by the neural network under state \(s_i\), similar to the onehot encoding scheme.
The evolution process of each EFS regressor in the regressor set is the same as previously mentioned in “Evolutionary feature synthesis” section, consisting of three major steps: feature generation, feature evaluation, feature selection.

Feature generation: New features are generated by randomly composing old features and new operators in this stage.

Feature evaluation: The importance of each feature is calculated by the fregression or the mutual information algorithm in this stage.

Feature selection: Features are ranked according to the feature importance after the feature evaluation process. Topranked features are selected to construct a new linear model. Model parameters are fitted by the least square method.
Figure 4 depicts the feature selection process based on the p value.
EFS ensembling
In a real environment, for a specific state s, action a should be generated by the policy model. For EFS, the action chosen by EFS is corresponding to the action label of the regressor, which has the most significant predicted value. Formally, action a is defined by the following formula.
The action generation process is presented in the Fig. 5. Due to the action A has the largest predicted value, the action A is chosen as the real action at the current state.
Operator optimization
Domain knowledge is useful in the reinforcement learning domain. Some specific control operators may be more expressive than other operators in a specific problem. Therefore, incorporating domain knowledge to select operators can dramatically reduce the search space and consequently obtain much better results.
However, operator selection based on the domain knowledge is heavily dependent on the domain experts. In many cases, domain experts are lacked. Sometimes, even with domain experts exist, the wellperformed operator may be counterintuitive. Therefore, automatically selecting the best operators are preferred in the operator optimization process.
In this paper, we use a binary encoding based genetic algorithm to select the best operators for a specific problem. First, in the initialization stage, we predefine an operator set O contains all potential useful operators, such as \(O=\{+,,*,/\}\). Then, binary chromosomes P with the length of O each chromosome are created, and O is the number of operators in the operator set O. Each chromosome \(p \in P\) represents an operator selection scheme \(O' \subseteq O\).
In the reproduction process, new chromosomes are generated by uniform random twopoint crossover and onepoint mutation [13]. In the evaluation process, for each operator selection scheme \(O'\), an EFS with less training iterations and evaluating episodes is used here to assessing the performance of the operator subset. In our experiments, we use twofifths of the normal training iterations and onefifth of the normal evaluating episodes as parameters. After that, the fitness \(T_{O'}\) is determined by the score of the final regressor on the specific environment. In the selection process, chromosomes are selected by the tournament selection algorithm [13]. The overall process is repeated until N iterations have reached. Algorithm 3 presents the whole process of the operator optimization algorithm.
Policy simplification
In the interpretable machine learning domain, multiobjective models are preferred by practitioners to achieve the balance between the tradeoff of complexity and performance. Due to the unique nature of the EFS algorithm, with the policy simplification method, a series of models with different complexity and performance can be obtained from the best policy model trivially.
Suppose that the number of features F is the complexity indicator of the model. The ultimate goal of the policy simplification is to derive a series of policy models with a different number of features and reasonable rewards. Therefore, for the best model evolved by the aforementioned algorithm, we can rank the importance of those features in that model by the pvalue. Then progressively integrating those features to compose a new feature set \(F' \subset F\) by the rank value. Finally, a series of feature sets compose a series of linear models R. It should be noted that for each model f, the model parameters are determined by the training data D through the least square method. Formally, the whole process of policy simplification is presented in Algorithm 4.
Experiment
Experiment settings

Baseline method: GP, Neural network, and common interpretable machine learning algorithms are used as baseline methods. The GP method derives the policy form the neural network by using the treebased genetic programming algorithm [17]. The neural network is used as the baseline method because the derived policy can even surpass the original neural network in the following experiments. For the interpretable machine learning algorithms, we use linear regression, decision tree, and Knearest neighbor algorithms in the following experiments. In the machine learning community, these algorithms are widely considered as the high interpretable models [5].

Benchmark datasets: We compare various methods on the CartPole environment, the Acrobot environment, the MountainCar environment, and the Industrial Benchmark environment.

Evaluation metrics: Mean square error (MSE) and real environment reward are used to assess the performance of different algorithms. The mean square error is defined as
$$\begin{aligned} \mathrm{MSE}(f)=\frac{\sum _{a \in A} (f_a(x)y)^2}{SA}, \end{aligned}$$(7)where S represents the size of training data, x represents the state data \(s \in S\), y represents \(I_a(s)\). The real environment reward is defined by the specific problem.

Complexity metrics: Tree height and weighted number of leaf nodes are used as complexity metrics to appraise the interpretability of GP and EFS. Similar to the previous study [35], we prefer a flatter tree with fewer layers than a deeper tree. Therefore, the weighted number of leaf nodes is calculated as \(\sum _{l\in L} \mathrm{depth}(l)\), where \(\mathrm{depth}(l)\) represents the depth of the leaf node l. Due to each model is composed of several regressors, we sum up the tree height and weighted number of leaf nodes as the complexity value of a specific model.
Experiment environment
CartPole
In the cartpole problem [31], an unactuated pole is connected to a car among a frictionless track. The goal is to prevent the pole fall over from the cart. In each step, two actions \(\{\mathrm{left,right}\}\) can be chosen to apply a force from the left of the cart or from the right of the cart. The reward r(s) is 1 at each step until the game over. In our experiment, 1001 trials are selected as the upper limit trials because some interesting phenomenons will emerge when longer trials are used.
Acrobot
In the acrobot problem [30], two links are connected by a joint. The target of this problem is to swing the end of the lower link to reach a specific height. In each step, three actions \(\{\mathrm{left,neutral,right}\}\) can be chosen to apply a force to the joint. The reward r(s) is \(1\) at each step until the joint reaches target height.
MountainCar
In the mountain car problem [23], an underpowered vehicle is situated in a valley at the position \(p_0 \in (0.6,0.4)\) with speed \(v_0=0\). The destination is located at \(p_d=0.5\), and the target of the problem is to reach the target location with the lowest cost. In each step, three actions \(\{\mathrm{left,neutral,right}\}\) can be chosen to execute, and the reward r(s) at state s is \(1\) until it reaches the target location.
Industrial benchmark
In the Industrial Benchmark problem [10], we control a simulation environment through 27 kinds of operations. The target of this problem is to minimize the cost, which is defined as three times the fatigue f, plus the consumption c, i.e., \(\mathrm{cost}=3 \times f+c\). Each episode is composed of 1000 steps, the final reward is calculated as the sum of costs in the episode. The details of the simulation environment are presented in [10].
Parameter settings
Due to the randomness of the evolutionary algorithm, experiments of GP and EFS are independently repeated 30 times, and at each time, the final model plays 100 episodes of games consecutively. For GP and EFS, 50 iterations of evolution are performed. The max height of GP is limited as 25. The number of constructed features of EFS is restricted as 10, and the max size of each feature is set as 8. The fregression is used as the feature importance evaluation approach. The details of the experiment parameters are presented in Table 1.
We try to cover a broad range of arithmetic operators to describe the control policy in the experiment. operator \(\in \{ +,,*,/,\mathrm{negative},\%,\mathrm{max,min,logistic}(x), \mathrm{log}_2{(x)}, \)ln(x)\(,\mathrm{log}_{10}{(x)}, \sqrt{x},x^3,x^2,x^{1/3},\sin (x),\cos (x),x\}\) operation optimization has been performed. For the cartpole problem, the selected operators are \(\{+,,*,/,\%, \mathrm{logistic}(x), \mathrm{ln}(x), \mathrm{log}_{2}(x), \mathrm{log}_{10}{(x)},{x}^2\}\). For the mountain car problem, the selected operators are \(\{+,,*,/, \mathrm{ln}(x), \mathrm{log}_{2}(x), \sqrt{x},\root 3 \of {x}, \sin (x)\}\). For the acrobot problem, the selected operators are \(\{+,,*,/,\mathrm{max}, \mathrm{ln}(x), {x}^2,\sqrt{x},\sin (x), \cos (x)\}\). For the industrial benchmark problem, the selected operators are \(\{+,,*,/,\%,\mathrm{min}, \mathrm{log}_{2}(x), {x}^2,\sin (x),x\}\). In the experiment, the standard logistic function is used as the logistic function, which is defined below.
Due to some operators have restrictions on the input values, we take some protective measures to avoid accident results. For logarithmic operators, we wrap these operators by the following function.
For the division operator, we avoid division by zero exception by using the following function. The modulo operator is also wrapping by a similar function.
The neural networks are pretrained by Qlearning [22] and policy gradient [32]. For the mountain car and the cartpole problem, a network with two fully connected hidden layers with 256 neurons is used. For the acrobot problem, a network with one hidden layer with 32 neurons is used. For the industrial benchmark problem, we use 5 hidden layers with 64 neurons, followed by a layer with 32 neurons. For all problems, the rectified linear unit (ReLU) [8] is used as the activation function, and Adam optimizer is used as the learning rate optimizer [16]. For the mountain car problem and the cartpole problem, the neural network trained until the target reward has reached with 100 consecutive trials. For the acrobot problem and the industrial benchmark problem, the neural network trained until the maximum iteration number is reached. The detailed parameters are presented in Table 2.
For common interpretable machine learning algorithms, we use the default parameters implemented in the Scikitlearn package [26]. The default parameters do not restrict the height of the decision tree, and the nearest neighbors of the Knearest neighbor algorithm is set as five.
Experimental result
Figure 6 presents the training error of four problems. The experimental results show that the training error of EFS is similar to GP. However, the real environment rewards of four problems presented in Fig. 7 show that EFS outperforms GP significantly. It should be noted that the performance of EFS exceeds our expectations, which outperforms the neural network on all problems, this may be caused by that operator selection has selected proper operators to describe the control policy of these problems. The numerical results of different algorithms are presented in Table 3.
From the numerical results, we find that EFS is not only superior to genetic programming but also surpass traditional interpretable machine learning models. For the simplest model, linear regression, we find that it is hard to mimic the behavior of the deep neural network and performs poorly on the Acrobot and MountainCar problem. However, the linear regression model performs well on the Cartpole problem. The reason behind this phenomenon is that the low complexity of the linear model may lead it more robust in the real environment. For the decision tree and the KNN, we find that the performances of the two algorithms are very similar to each other. These two models are accurately memoried the behavior of the neural network. However, they are confined by the performance of the neural network. In contrast to these algorithms, EFS significantly outperforms these algorithms by using expressive features.
Influence of operator optimization
Figure 8 presents the training error of four problems with different operator optimization strategies. Figure 9 presents the real environment rewards of four problems with different operator optimization strategies. From these results, we find that even though the operator optimization reduce the fitting accuracy, especially on the CartPole problem. The final rewards of the policy composed by optimized operators outperform rewards of the policy without optimized operators. The discrepancy between the fitting accuracy and final rewards shows the importance of suitable operators for different problems. In some cases, even if some inappropriate operators achieve good accuracy in the training process, it may not generalize well. The reason might be that it does not capture the true pattern of the control rules. The numerical results of two different strategies are presented in Table 4.
Result of policy simplification
The results of policy simplification are depicted in Fig. 10, and the numerical results are presented in Table 5. The complexity represents the number of evolved features in the linear model, and complexity with 0 means that only original features are contained in the model. Four problems exhibit different characteristics. For the cartpole and the industrial benchmark problem, lower complexity models attain much reward than midsize models. For the mountain car problem, the original linear model (complexity 0) needs a mean cost of 195 to solve this task. However, adding only one evolved feature, then the model can achieve a reasonable result with an average reward of \(119\). The acrobot problem behaves similar to the mountain car problem, a linear model with at least four evolved features can achieve a reasonable result.
Structural complexity analysis
Tables 6 and 7 presents the depth and weighted number of leaf nodes of different algorithms. It is necessary to emphasize that the complexity of a specific model is the sum of the complexity of regressors in the model. Table 6 suggests that policy derived by EFS is significantly shallower than GP in the Cartpole, Acrobot, and IndustrialBenchmark environments. Table 7 reveals that the weighted number of leaf nodes of EFS is significantly lesser than GP in the Cartpole environment and similar to GP in other environments. From these results, we can conclude that the explainability of EFS is comparable to GP.
Time complexity analysis
Table 8 presents the training time of different algorithms of four problems. Experimental results show that EFS requires more time than GP on all problems. The time used by EFS is 2.6815.87 times of the running time used by GP. In the future, the more efficient feature importance evaluation method is worth to be explored to achieve better performance.
Explainable result
We obtain a controller from the experiment mentioned above. For brevity, the value is rounded to three decimal places. The control function is described below.
\(X={[x_0,x_1,x_2,x_3]}\) represents the state value, and the corresponding action a is determined by the result of f(X)
We convert this function into a Python code fragment and evaluate it on the cartpole environment implemented by OpenAIGym [4]. We conduct 100 consecutive experiments, and it achieves an average reward of 993. This experimental result is similar to the previous result, and the decimal round operation has not damaged the function performance.
Besides the controller example of the Cartpole problem, we also present an EFS controller for the industrial benchmark problem in Fig. 11, which obtains an average reward of \(1981\) in 100 consecutive trials. We only show five actions due to only those five actions are used by the trained neural network, and the training data generated by the neural network only contain those five actions. Therefore, the EFS controller will not try to use other actions.
Conclusion
In this paper, we propose a novel method to derive interpretable policy from the neural network by evolutionary feature synthesis (EFS). We first propose a framework derives interpretable policy from the behavior data of the neural network. Then an EA based operator optimization is designed to automatically select the best operators for a specific problem to reduce the search space and enhance the performance. A policy simplification algorithm is also proposed to balance the tradeoff between the performance and the interpretability. We study the performance of EFS based on four control problems. The experimental results show that the policy derived by EFS outperforms the policy derived by treebased genetic programming (GP) and common interpretable machine learning algorithms, with comparable interpretability to the genetic programming algorithm.
For future work, it is deserved to investigate the performance of policy derived by EFS on more complicated problems, such as computer games. It is also worth to explore that combining more sophisticated domain operators to achieve better performance.
References
Alibekov E, Kubalík J, Babuska R (2016) Symbolic method for deriving policy in reinforcement learning. In: 55th IEEE conference on decision and control, CDC 2016. IEEE, pp 2789–2795
Alibekov E, Kubalík J, Babuska R (2018) Policy derivation methods for criticonly reinforcement learning in continuous spaces. Eng Appl Artif Intell 69:178–187
Arnaldo I, O’Reilly U, Veeramachaneni K (2015) Building predictive models via feature synthesis. In: Proceedings of the genetic and evolutionary computation conference, GECCO 2015. ACM, pp 983–990
Brockman G, Cheung V, Pettersson L et al (2016) Openai gym. CoRR. arXiv:1606.01540
Cayamcela MEM, Lee H, Lim W (2019) Machine learning for 5g/b5g mobile and wireless communications: potential, limitations, and future directions. IEEE Access 7:137184–137206
Derner E, Kubalík J, Babuska R (2018) Datadriven construction of symbolic process models for reinforcement learning. In: 2018 IEEE international conference on robotics and automation, ICRA 2018. IEEE, pp 1–8
Derner E, Kubalík J, Babuska R (2018) Reinforcement learning with symbolic inputoutput models. In: 2018 IEEE/RSJ international conference on intelligent robots and systems, IROS 2018. IEEE, pp 3004–3009
Glorot X, Bordes A, Bengio Y (2011) Deep sparse rectifier neural networks. In: Proceedings of the fourteenth international conference on artificial intelligence and statistics, AISTATS 2011, JMLR proceedings, vol 15, pp 315–323. JMLR.org
van Hasselt H, Guez A, Silver D (2016) Deep reinforcement learning with double qlearning. In: Proceedings of the thirtieth AAAI conference on artificial intelligence, AAAI 2016. AAAI Press, pp 2094–2100
Hein D, Depeweg S, Tokic M et al (2017) A benchmark environment motivated by industrial control problems. In: 2017 IEEE symposium series on computational intelligence, SSCI 2017. IEEE, pp 1–8
Hein D, Udluft S, Runkler TA (2018) Interpretable policies for reinforcement learning by genetic programming. Eng Appl Artif Intell 76:158–169
Hirasawa K, Okubo M, Katagiri H, Hu J, Murata J (2001) Comparison between genetic network programming (gnp) and genetic programming (gp). In: Proceedings of the 2001 congress on evolutionary computation, CEC 2001, vol 2. IEEE, pp 1276–1282
Hussain A, Muhammad YS (2019) Tradeoff between exploration and exploitation with genetic algorithm using a novel selection operator. In: Complex & intelligent systems, pp 1–14
Katagiri H, Hirasama K, Hu J (2000) Genetic network programming—application to intelligent agents. In: Proceedings of the IEEE international conference on systems, man & cybernetics: “cybernetics evolving to systems, humans, organizations, and their complex interactions”, SMC 2000. IEEE, pp 3829–3834
Kim S, Ha J, Kim H, Zhang B (2019) Bayesian evolutionary hypernetworks for interpretable learning from highdimensional data. Appl Soft Comput 81:104577. https://doi.org/10.1016/j.asoc.2019.05.004
Kingma DP, Ba J (2015) Adam: a method for stochastic optimization. In: 3rd international conference on learning representations, ICLR 2015
Koza JR, Koza JR (1992) Genetic programming: on the programming of computers by means of natural selection, vol 1. MIT Press, Cambridge
Kubalík J, Zegklitz J, Derner E, Babuska R (2019) Symbolic regression methods for reinforcement learning. CoRR. arXiv:1903.09688
Mabu S, Hirasawa K, Hu J (2004) Genetic network programming with reinforcement learning and its performance evaluation. In: Genetic and evolutionary computation—GECCO 2004, Lecture notes in computer science, vol 3103. Springer, pp 710–711
Mabu S, Hirasawa K, Hu J, Murata J (2002) Online learning of genetic network programming (gnp). In: Proceedings of the 2002 congress on evolutionary computation, CEC 2002, vol 1. IEEE, pp 321–326
Mnih V, Badia AP, Mirza M et al (2016) Asynchronous methods for deep reinforcement learning. In: Proceedings of the 33nd international conference on machine learning, ICML 2016, JMLR workshop and conference proceedings, vol 48, pp 1928–1937. JMLR.org
Mnih V, Kavukcuoglu K, Silver D et al (2015) Humanlevel control through deep reinforcement learning. Nature 518(7540):529
Moore AW (1990) Efficient memorybased learning for robot control. Tech. rep., University of Cambridge, Computer Laboratory
Nguyen S, Mei Y, Zhang M (2017) Genetic programming for production scheduling: a survey with a unified framework. Complex Intell Syst 3(1):41–66
Onderwater M, Bhulai S, van der Mei R (2016) Value function discovery in markov decision processes with evolutionary algorithms. IEEE Trans Syst Man Cybern Syst 46(9):1190–1201
Pedregosa F, Varoquaux G, Gramfort A et al (2011) Scikitlearn: machine learning in python. J Mach Learn Res 12(Oct):2825–2830
Rudin C (2019) Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nat Mach Intell 1(5):206–215
Shimooka H, Fujimoto Y (1998) Generating equations with genetic programming for control of a movable inverted pendulum. In: Simulated evolution and learning, second AsiaPacific conference on simulated evolution and learning, SEAL 98, Lecture notes in computer science, vol 1585. Springer, pp 179–186
Silver D, Schrittwieser J, Simonyan K et al (2017) Mastering the game of go without human knowledge. Nature 550(7676):354
Sutton RS (1995) Generalization in reinforcement learning: successful examples using sparse coarse coding. In: Advances in neural information processing systems, vol 8, NIPS 1995. MIT Press, pp 1038–1044
Sutton RS, Barto AG (2018) Reinforcement learning: an introduction. MIT Press, Cambridge
Sutton RS, McAllester DA, Singh SP, Mansour Y (1999) Policy gradient methods for reinforcement learning with function approximation. In: Advances in neural information processing systems, vol 12, NIPS 1999. The MIT Press, pp 1057–1063
Vellido A (2019) The importance of interpretability and visualization in machine learning for applications in medicine and health care. Neural Comput Appl 1–15. https://doi.org/10.1007/s0052101904051w
Verma L, Srivastava S, Negi P (2018) An intelligent noninvasive model for coronary artery disease detection. Complex Intell Syst 4(1):11–18
Vladislavleva EJ, Smits GF, Den Hertog D (2008) Order of nonlinearity as a complexity measure for models generated by symbolic regression via pareto genetic programming. IEEE Trans Evolut Comput 13(2):333–349
Wang Z, Schaul T, Hessel M et al (2016) Dueling network architectures for deep reinforcement learning. In: Proceedings of the 33nd international conference on machine learning, ICML 2016, JMLR workshop and conference proceedings, vol 48, pp 1995–2003. JMLR.org
Acknowledgements
This work is supported by the Science and Technology Commission of Shanghai Municipality (No. 19511120600) and the National Nature Science Foundation of China (Nos. 61773296, 61673180).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Zhang, H., Zhou, A. & Lin, X. Interpretable policy derivation for reinforcement learning based on evolutionary feature synthesis. Complex Intell. Syst. 6, 741–753 (2020). https://doi.org/10.1007/s4074702000175y
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s4074702000175y