Frederik Schubert

Frederik Schubert, Carolin Benjamins, Sebastian Döhler et al.

The goal of Unsupervised Reinforcement Learning (URL) is to find a reward-agnostic prior policy on a task domain, such that the sample-efficiency on supervised downstream tasks is improved. Although agents initialized with such a prior policy can achieve a significantly higher reward with fewer samples when finetuned on the downstream task, it is still an open question how an optimal pretrained prior policy can be achieved in practice. In this work, we present POLTER (Policy Trajectory Ensemble Regularization) - a general method to regularize the pretraining that can be applied to any URL algorithm and is especially useful on data- and knowledge-based URL algorithms. It utilizes an ensemble of policies that are discovered during pretraining and moves the policy of the URL algorithm closer to its optimal prior. Our method is based on a theoretical framework, and we analyze its practical effects on a white-box benchmark, allowing us to study POLTER with full control. In our main experiments, we evaluate POLTER on the Unsupervised Reinforcement Learning Benchmark (URLB), which consists of 12 tasks in 3 domains. We demonstrate the generality of our approach by improving the performance of a diverse set of data- and knowledge-based URL algorithms by 19% on average and up to 40% in the best case. Under a fair comparison with tuned baselines and tuned POLTER, we establish a new state-of-the-art for model-free methods on the URLB.

Carolin Benjamins, Theresa Eimer, Frederik Schubert et al.

While Reinforcement Learning has made great strides towards solving ever more complicated tasks, many algorithms are still brittle to even slight changes in their environment. This is a limiting factor for real-world applications of RL. Although the research community continuously aims at improving both robustness and generalization of RL algorithms, unfortunately it still lacks an open-source set of well-defined benchmark problems based on a consistent theoretical framework, which allows comparing different approaches in a fair, reliable and reproducibleway. To fill this gap, we propose CARL, a collection of well-known RL environments extended to contextual RL problems to study generalization. We show the urgent need of such benchmarks by demonstrating that even simple toy environments become challenging for commonly used approaches if different contextual instances of this task have to be considered. Furthermore, CARL allows us to provide first evidence that disentangling representation learning of the states from the policy learning with the context facilitates better generalization. By providing variations of diverse benchmarks from classic control, physical simulations, games and a real-world application of RNA design, CARL will allow the community to derive many more such insights on a solid empirical foundation.

Christoph Reinders, Frederik Schubert, Bodo Rosenhahn

Training deep neural networks requires datasets with a large number of annotated examples. The collection and annotation of these datasets is not only extremely expensive but also faces legal and privacy problems. These factors are a significant limitation for many real-world applications. To address this, we introduce HydraMix, a novel architecture that generates new image compositions by mixing multiple different images from the same class. HydraMix learns the fusion of the content of various images guided by a segmentation-based mixing mask in feature space and is optimized via a combination of unsupervised and adversarial training. Our data augmentation scheme allows the creation of models trained from scratch on very small datasets. We conduct extensive experiments on ciFAIR-10, STL-10, and ciFAIR-100. Additionally, we introduce a novel text-image metric to assess the generality of the augmented datasets. Our results show that HydraMix outperforms existing state-of-the-art methods for image classification on small datasets.

Yannik Mahlau, Frederik Schubert, Bodo Rosenhahn

The combination of self-play and planning has achieved great successes in sequential games, for instance in Chess and Go. However, adapting algorithms such as AlphaZero to simultaneous games poses a new challenge. In these games, missing information about concurrent actions of other agents is a limiting factor as they may select different Nash equilibria or do not play optimally at all. Thus, it is vital to model the behavior of the other agents when interacting with them in simultaneous games. To this end, we propose Albatross: AlphaZero for Learning Bounded-rational Agents and Temperature-based Response Optimization using Simulated Self-play. Albatross learns to play the novel equilibrium concept of a Smooth Best Response Logit Equilibrium (SBRLE), which enables cooperation and competition with agents of any playing strength. We perform an extensive evaluation of Albatross on a set of cooperative and competitive simultaneous perfect-information games. In contrast to AlphaZero, Albatross is able to exploit weak agents in the competitive game of Battlesnake. Additionally, it yields an improvement of 37.6% compared to previous state of the art in the cooperative Overcooked benchmark.

Yannik Mahlau, Maximilian Schier, Christoph Reinders et al.

Inverse design of photonic integrated circuits (PICs) has traditionally relied on gradientbased optimization. However, this approach is prone to end up in local minima, which results in suboptimal design functionality. As interest in PICs increases due to their potential for addressing modern hardware demands through optical computing, more adaptive optimization algorithms are needed. We present a reinforcement learning (RL) environment as well as multi-agent RL algorithms for the design of PICs. By discretizing the design space into a grid, we formulate the design task as an optimization problem with thousands of binary variables. We consider multiple two- and three-dimensional design tasks that represent PIC components for an optical computing system. By decomposing the design space into thousands of individual agents, our algorithms are able to optimize designs with only a few thousand environment samples. They outperform previous state-of-the-art gradient-based optimization in both twoand three-dimensional design tasks. Our work may also serve as a benchmark for further exploration of sample-efficient RL for inverse design in photonics.

Christoph Reinders, Frederik Schubert, Bodo Rosenhahn

Deep convolutional neural networks require large amounts of labeled data samples. For many real-world applications, this is a major limitation which is commonly treated by augmentation methods. In this work, we address the problem of learning deep neural networks on small datasets. Our proposed architecture called ChimeraMix learns a data augmentation by generating compositions of instances. The generative model encodes images in pairs, combines the features guided by a mask, and creates new samples. For evaluation, all methods are trained from scratch without any additional data. Several experiments on benchmark datasets, e.g. ciFAIR-10, STL-10, and ciFAIR-100, demonstrate the superior performance of ChimeraMix compared to current state-of-the-art methods for classification on small datasets.

Carolin Benjamins, Theresa Eimer, Frederik Schubert et al.

While Reinforcement Learning ( RL) has made great strides towards solving increasingly complicated problems, many algorithms are still brittle to even slight environmental changes. Contextual Reinforcement Learning (cRL) provides a framework to model such changes in a principled manner, thereby enabling flexible, precise and interpretable task specification and generation. Our goal is to show how the framework of cRL contributes to improving zero-shot generalization in RL through meaningful benchmarks and structured reasoning about generalization tasks. We confirm the insight that optimal behavior in cRL requires context information, as in other related areas of partial observability. To empirically validate this in the cRL framework, we provide various context-extended versions of common RL environments. They are part of the first benchmark library, CARL, designed for generalization based on cRL extensions of popular benchmarks, which we propose as a testbed to further study general agents. We show that in the contextual setting, even simple RL environments become challenging - and that naive solutions are not enough to generalize across complex context spaces.

Maren Awiszus, Frederik Schubert, Bodo Rosenhahn

This work introduces World-GAN, the first method to perform data-driven Procedural Content Generation via Machine Learning in Minecraft from a single example. Based on a 3D Generative Adversarial Network (GAN) architecture, we are able to create arbitrarily sized world snippets from a given sample. We evaluate our approach on creations from the community as well as structures generated with the Minecraft World Generator. Our method is motivated by the dense representations used in Natural Language Processing (NLP) introduced with word2vec [1]. The proposed block2vec representations make World-GAN independent from the number of different blocks, which can vary a lot in Minecraft, and enable the generation of larger levels. Finally, we demonstrate that changing this new representation space allows us to change the generated style of an already trained generator. World-GAN enables its users to generate Minecraft worlds based on parts of their creations.

Frederik Schubert, Theresa Eimer, Bodo Rosenhahn et al.

The use of Reinforcement Learning (RL) agents in practical applications requires the consideration of suboptimal outcomes, depending on the familiarity of the agent with its environment. This is especially important in safety-critical environments, where errors can lead to high costs or damage. In distributional RL, the risk-sensitivity can be controlled via different distortion measures of the estimated return distribution. However, these distortion functions require an estimate of the risk level, which is difficult to obtain and depends on the current state. In this work, we demonstrate the suboptimality of a static risk level estimation and propose a method to dynamically select risk levels at each environment step. Our method ARA (Automatic Risk Adaptation) estimates the appropriate risk level in both known and unknown environments using a Random Network Distillation error. We show reduced failure rates by up to a factor of 7 and improved generalization performance by up to 14% compared to both risk-aware and risk-agnostic agents in several locomotion environments.

Maren Awiszus, Frederik Schubert, Bodo Rosenhahn

In this work, we present TOAD-GAN (Token-based One-shot Arbitrary Dimension Generative Adversarial Network), a novel Procedural Content Generation (PCG) algorithm that generates token-based video game levels. TOAD-GAN follows the SinGAN architecture and can be trained using only one example. We demonstrate its application for Super Mario Bros. levels and are able to generate new levels of similar style in arbitrary sizes. We achieve state-of-the-art results in modeling the patterns of the training level and provide a comparison with different baselines under several metrics. Additionally, we present an extension of the method that allows the user to control the generation process of certain token structures to ensure a coherent global level layout. We provide this tool to the community to spur further research by publishing our source code.