Reactive Exploration to Cope with Non-Stationarity in Lifelong Reinforcement Learning

In lifelong learning, an agent learns throughout its entire life without resets, in a constantly changing environment, as we humans do. Consequently, lifelong learning comes with a plethora of research problems such as continual domain shifts, which result in non-stationary rewards and environment dynamics. These non-stationarities are difficult to detect and cope with due to their continuous nature. Therefore, exploration strategies and learning methods are required that are capable of tracking the steady domain shifts, and adapting to them. We propose Reactive Exploration to track and react to continual domain shifts in lifelong reinforcement learning, and to update the policy correspondingly. To this end, we conduct experiments in order to investigate different exploration strategies. We empirically show that representatives of the policy-gradient family are better suited for lifelong learning, as they adapt more quickly to distribution shifts than Q-learning. Thereby, policy-gradient methods profit the most from Reactive Exploration and show good results in lifelong learning with continual domain shifts. Our code is available at: https://github.com/ml-jku/reactive-exploration.

Few-Shot Learning by Dimensionality Reduction in Gradient Space

We introduce SubGD, a novel few-shot learning method which is based on the recent finding that stochastic gradient descent updates tend to live in a low-dimensional parameter subspace. In experimental and theoretical analyses, we show that models confined to a suitable predefined subspace generalize well for few-shot learning. A suitable subspace fulfills three criteria across the given tasks: it (a) allows to reduce the training error by gradient flow, (b) leads to models that generalize well, and (c) can be identified by stochastic gradient descent. SubGD identifies these subspaces from an eigendecomposition of the auto-correlation matrix of update directions across different tasks. Demonstrably, we can identify low-dimensional suitable subspaces for few-shot learning of dynamical systems, which have varying properties described by one or few parameters of the analytical system description. Such systems are ubiquitous among real-world applications in science and engineering. We experimentally corroborate the advantages of SubGD on three distinct dynamical systems problem settings, significantly outperforming popular few-shot learning methods both in terms of sample efficiency and performance.

Entangled Residual Mappings

Residual mappings have been shown to perform representation learning in the first layers and iterative feature refinement in higher layers. This interplay, combined with their stabilizing effect on the gradient norms, enables them to train very deep networks. In this paper, we take a step further and introduce entangled residual mappings to generalize the structure of the residual connections and evaluate their role in iterative learning representations. An entangled residual mapping replaces the identity skip connections with specialized entangled mappings such as orthogonal, sparse, and structural correlation matrices that share key attributes (eigenvalues, structure, and Jacobian norm) with identity mappings. We show that while entangled mappings can preserve the iterative refinement of features across various deep models, they influence the representation learning process in convolutional networks differently than attention-based models and recurrent neural networks. In general, we find that for CNNs and Vision Transformers entangled sparse mapping can help generalization while orthogonal mappings hurt performance. For recurrent networks, orthogonal residual mappings form an inductive bias for time-variant sequences, which degrades accuracy on time-invariant tasks.

Hopular: Modern Hopfield Networks for Tabular Data

While Deep Learning excels in structured data as encountered in vision and natural language processing, it failed to meet its expectations on tabular data. For tabular data, Support Vector Machines (SVMs), Random Forests, and Gradient Boosting are the best performing techniques with Gradient Boosting in the lead. Recently, we saw a surge of Deep Learning methods that were tailored to tabular data but still underperform compared to Gradient Boosting on small-sized datasets. We suggest "Hopular", a novel Deep Learning architecture for medium- and small-sized datasets, where each layer is equipped with continuous modern Hopfield networks. The modern Hopfield networks use stored data to identify feature-feature, feature-target, and sample-sample dependencies. Hopular's novelty is that every layer can directly access the original input as well as the whole training set via stored data in the Hopfield networks. Therefore, Hopular can step-wise update its current model and the resulting prediction at every layer like standard iterative learning algorithms. In experiments on small-sized tabular datasets with less than 1,000 samples, Hopular surpasses Gradient Boosting, Random Forests, SVMs, and in particular several Deep Learning methods. In experiments on medium-sized tabular data with about 10,000 samples, Hopular outperforms XGBoost, CatBoost, LightGBM and a state-of-the art Deep Learning method designed for tabular data. Thus, Hopular is a strong alternative to these methods on tabular data.

History Compression via Language Models in Reinforcement Learning

In a partially observable Markov decision process (POMDP), an agent typically uses a representation of the past to approximate the underlying MDP. We propose to utilize a frozen Pretrained Language Transformer (PLT) for history representation and compression to improve sample efficiency. To avoid training of the Transformer, we introduce FrozenHopfield, which automatically associates observations with original token embeddings. To form these associations, a modern Hopfield network stores the original token embeddings, which are retrieved by queries that are obtained by a random but fixed projection of observations. Our new method, HELM, enables actor-critic network architectures that contain a pretrained language Transformer for history representation as a memory module. Since a representation of the past need not be learned, HELM is much more sample efficient than competitors. On Minigrid and Procgen environments HELM achieves new state-of-the-art results. Our code is available at https://github.com/ml-jku/helm.

Traffic4cast at NeurIPS 2021 -- Temporal and Spatial Few-Shot Transfer Learning in Gridded Geo-Spatial Processes

The IARAI Traffic4cast competitions at NeurIPS 2019 and 2020 showed that neural networks can successfully predict future traffic conditions 1 hour into the future on simply aggregated GPS probe data in time and space bins. We thus reinterpreted the challenge of forecasting traffic conditions as a movie completion task. U-Nets proved to be the winning architecture, demonstrating an ability to extract relevant features in this complex real-world geo-spatial process. Building on the previous competitions, Traffic4cast 2021 now focuses on the question of model robustness and generalizability across time and space. Moving from one city to an entirely different city, or moving from pre-COVID times to times after COVID hit the world thus introduces a clear domain shift. We thus, for the first time, release data featuring such domain shifts. The competition now covers ten cities over 2 years, providing data compiled from over 10^12 GPS probe data. Winning solutions captured traffic dynamics sufficiently well to even cope with these complex domain shifts. Surprisingly, this seemed to require only the previous 1h traffic dynamic history and static road graph as input.

Understanding the Effects of Dataset Characteristics on Offline Reinforcement Learning

In real world, affecting the environment by a weak policy can be expensive or very risky, therefore hampers real world applications of reinforcement learning. Offline Reinforcement Learning (RL) can learn policies from a given dataset without interacting with the environment. However, the dataset is the only source of information for an Offline RL algorithm and determines the performance of the learned policy. We still lack studies on how dataset characteristics influence different Offline RL algorithms. Therefore, we conducted a comprehensive empirical analysis of how dataset characteristics effect the performance of Offline RL algorithms for discrete action environments. A dataset is characterized by two metrics: (1) the average dataset return measured by the Trajectory Quality (TQ) and (2) the coverage measured by the State-Action Coverage (SACo). We found that variants of the off-policy Deep Q-Network family require datasets with high SACo to perform well. Algorithms that constrain the learned policy towards the given dataset perform well for datasets with high TQ or SACo. For datasets with high TQ, Behavior Cloning outperforms or performs similarly to the best Offline RL algorithms.

CLOOB: Modern Hopfield Networks with InfoLOOB Outperform CLIP

Contrastive learning with the InfoNCE objective is exceptionally successful in various self-supervised learning tasks. Recently, the CLIP model yielded impressive results on zero-shot transfer learning when using InfoNCE for learning visual representations from natural language supervision. However, InfoNCE as a lower bound on the mutual information has been shown to perform poorly for high mutual information. In contrast, the InfoLOOB upper bound (leave one out bound) works well for high mutual information but suffers from large variance and instabilities. We introduce "Contrastive Leave One Out Boost" (CLOOB), where modern Hopfield networks boost learning with the InfoLOOB objective. Modern Hopfield networks replace the original embeddings by retrieved embeddings in the InfoLOOB objective. The retrieved embeddings give InfoLOOB two assets. Firstly, the retrieved embeddings stabilize InfoLOOB, since they are less noisy and more similar to one another than the original embeddings. Secondly, they are enriched by correlations, since the covariance structure of embeddings is reinforced through retrievals. We compare CLOOB to CLIP after learning on the Conceptual Captions and the YFCC dataset with respect to their zero-shot transfer learning performance on other datasets. CLOOB consistently outperforms CLIP at zero-shot transfer learning across all considered architectures and datasets.

Boundary Graph Neural Networks for 3D Simulations

The abundance of data has given machine learning huge momentum in natural sciences and engineering. However, the modeling of simulated physical processes remains difficult. A key problem in doing so is the correct handling of geometric boundaries. While triangularized geometric boundaries are very common in engineering applications, they are notoriously difficult to model by machine learning approaches due to their heterogeneity with respect to size and orientation. In this work, we introduce Boundary Graph Neural Networks (BGNNs), which dynamically modify graph structures to address boundary conditions. Boundary graph structures are constructed via modifying edges, augmenting node features, and dynamically inserting virtual nodes. The new BGNNs are tested on complex 3D granular flow processes of hoppers and rotating drums which are standard parts of industrial machinery. Using precise simulations that are obtained by an expensive and complex discrete element method, BGNNs are evaluated in terms of computational efficiency as well as prediction accuracy of particle flows and mixing entropies. Even if complex boundaries are present, BGNNs are able to accurately reproduce 3D granular flows within simulation uncertainties over hundreds of thousands of simulation timesteps, and most notably particles completely stay within the geometric objects without using handcrafted conditions or restrictions.

Learning 3D Granular Flow Simulations

Recently, the application of machine learning models has gained momentum in natural sciences and engineering, which is a natural fit due to the abundance of data in these fields. However, the modeling of physical processes from simulation data without first principle solutions remains difficult. Here, we present a Graph Neural Networks approach towards accurate modeling of complex 3D granular flow simulation processes created by the discrete element method LIGGGHTS and concentrate on simulations of physical systems found in real world applications like rotating drums and hoppers. We discuss how to implement Graph Neural Networks that deal with 3D objects, boundary conditions, particle - particle, and particle - boundary interactions such that an accurate modeling of relevant physical quantities is made possible. Finally, we compare the machine learning based trajectories to LIGGGHTS trajectories in terms of particle flows and mixing entropies.