Zero-shot Model-based Reinforcement Learning using Large Language Models

The emerging zero-shot capabilities of Large Language Models (LLMs) have led to their applications in areas extending well beyond natural language processing tasks. In reinforcement learning, while LLMs have been extensively used in text-based environments, their integration with continuous state spaces remains understudied. In this paper, we investigate how pre-trained LLMs can be leveraged to predict in context the dynamics of continuous Markov decision processes. We identify handling multivariate data and incorporating the control signal as key challenges that limit the potential of LLMs' deployment in this setup and propose Disentangled In-Context Learning (DICL) to address them. We present proof-of-concept applications in two reinforcement learning settings: model-based policy evaluation and data-augmented off-policy reinforcement learning, supported by theoretical analysis of the proposed methods. Our experiments further demonstrate that our approach produces well-calibrated uncertainty estimates. We release the code at https://github.com/abenechehab/dicl.

A call for embodied AI

We propose Embodied AI as the next fundamental step in the pursuit of Artificial General Intelligence, juxtaposing it against current AI advancements, particularly Large Language Models. We traverse the evolution of the embodiment concept across diverse fields - philosophy, psychology, neuroscience, and robotics - to highlight how EAI distinguishes itself from the classical paradigm of static learning. By broadening the scope of Embodied AI, we introduce a theoretical framework based on cognitive architectures, emphasizing perception, action, memory, and learning as essential components of an embodied agent. This framework is aligned with Friston's active inference principle, offering a comprehensive approach to EAI development. Despite the progress made in the field of AI, substantial challenges, such as the formulation of a novel AI learning theory and the innovation of advanced hardware, persist. Our discussion lays down a foundational guideline for future Embodied AI research. Highlighting the importance of creating Embodied AI agents capable of seamless communication, collaboration, and coexistence with humans and other intelligent entities within real-world environments, we aim to steer the AI community towards addressing the multifaceted challenges and seizing the opportunities that lie ahead in the quest for AGI.

A Multi-step Loss Function for Robust Learning of the Dynamics in Model-based Reinforcement Learning

In model-based reinforcement learning, most algorithms rely on simulating trajectories from one-step models of the dynamics learned on data. A critical challenge of this approach is the compounding of one-step prediction errors as the length of the trajectory grows. In this paper we tackle this issue by using a multi-step objective to train one-step models. Our objective is a weighted sum of the mean squared error (MSE) loss at various future horizons. We find that this new loss is particularly useful when the data is noisy (additive Gaussian noise in the observations), which is often the case in real-life environments. To support the multi-step loss, first we study its properties in two tractable cases: i) uni-dimensional linear system, and ii) two-parameter non-linear system. Second, we show in a variety of tasks (environments or datasets) that the models learned with this loss achieve a significant improvement in terms of the averaged R2-score on future prediction horizons. Finally, in the pure batch reinforcement learning setting, we demonstrate that one-step models serve as strong baselines when dynamics are deterministic, while multi-step models would be more advantageous in the presence of noise, highlighting the potential of our approach in real-world applications.

Deep autoregressive density nets vs neural ensembles for model-based offline reinforcement learning

We consider the problem of offline reinforcement learning where only a set of system transitions is made available for policy optimization. Following recent advances in the field, we consider a model-based reinforcement learning algorithm that infers the system dynamics from the available data and performs policy optimization on imaginary model rollouts. This approach is vulnerable to exploiting model errors which can lead to catastrophic failures on the real system. The standard solution is to rely on ensembles for uncertainty heuristics and to avoid exploiting the model where it is too uncertain. We challenge the popular belief that we must resort to ensembles by showing that better performance can be obtained with a single well-calibrated autoregressive model on the D4RL benchmark. We also analyze static metrics of model-learning and conclude on the important model properties for the final performance of the agent.

Multi-timestep models for Model-based Reinforcement Learning

In model-based reinforcement learning (MBRL), most algorithms rely on simulating trajectories from one-step dynamics models learned on data. A critical challenge of this approach is the compounding of one-step prediction errors as length of the trajectory grows. In this paper we tackle this issue by using a multi-timestep objective to train one-step models. Our objective is a weighted sum of a loss function (e.g., negative log-likelihood) at various future horizons. We explore and test a range of weights profiles. We find that exponentially decaying weights lead to models that significantly improve the long-horizon R2 score. This improvement is particularly noticeable when the models were evaluated on noisy data. Finally, using a soft actor-critic (SAC) agent in pure batch reinforcement learning (RL) and iterated batch RL scenarios, we found that our multi-timestep models outperform or match standard one-step models. This was especially evident in a noisy variant of the considered environment, highlighting the potential of our approach in real-world applications.

Guided Safe Shooting: model based reinforcement learning with safety constraints

In the last decade, reinforcement learning successfully solved complex control tasks and decision-making problems, like the Go board game. Yet, there are few success stories when it comes to deploying those algorithms to real-world scenarios. One of the reasons is the lack of guarantees when dealing with and avoiding unsafe states, a fundamental requirement in critical control engineering systems. In this paper, we introduce Guided Safe Shooting (GuSS), a model-based RL approach that can learn to control systems with minimal violations of the safety constraints. The model is learned on the data collected during the operation of the system in an iterated batch fashion, and is then used to plan for the best action to perform at each time step. We propose three different safe planners, one based on a simple random shooting strategy and two based on MAP-Elites, a more advanced divergent-search algorithm. Experiments show that these planners help the learning agent avoid unsafe situations while maximally exploring the state space, a necessary aspect when learning an accurate model of the system. Furthermore, compared to model-free approaches, learning a model allows GuSS reducing the number of interactions with the real-system while still reaching high rewards, a fundamental requirement when handling engineering systems.

Knothe-Rosenblatt transport for Unsupervised Domain Adaptation

Unsupervised domain adaptation (UDA) aims at exploiting related but different data sources to tackle a common task in a target domain. UDA remains a central yet challenging problem in machine learning. In this paper, we present an approach tailored to moderate-dimensional tabular problems which are hugely important in industrial applications and less well-served by the plethora of methods designed for image and language data. Knothe-Rosenblatt Domain Adaptation (KRDA) is based on the Knothe-Rosenblatt transport: we exploit autoregressive density estimation algorithms to accurately model the different sources by an autoregressive model using a mixture of Gaussians. KRDA then takes advantage of the triangularity of the autoregressive models to build an explicit mapping of the source samples into the target domain. We show that the transfer map built by KRDA preserves each component quantiles of the observations, hence aligning the representations of the different data sets in the same target domain. Finally, we show that KRDA has state-of-the-art performance on both synthetic and real world UDA problems.

Model-based micro-data reinforcement learning: what are the crucial model properties and which model to choose?

We contribute to micro-data model-based reinforcement learning (MBRL) by rigorously comparing popular generative models using a fixed (random shooting) control agent. We find that on an environment that requires multimodal posterior predictives, mixture density nets outperform all other models by a large margin. When multimodality is not required, our surprising finding is that we do not need probabilistic posterior predictives: deterministic models are on par, in fact they consistently (although non-significantly) outperform their probabilistic counterparts. We also found that heteroscedasticity at training time, perhaps acting as a regularizer, improves predictions at longer horizons. At the methodological side, we design metrics and an experimental protocol which can be used to evaluate the various models, predicting their asymptotic performance when using them on the control problem. Using this framework, we improve the state-of-the-art sample complexity of MBRL on Acrobot by two to four folds, using an aggressive training schedule which is outside of the hyperparameter interval usually considered

Tropical Cyclone Track Forecasting using Fused Deep Learning from Aligned Reanalysis Data

The forecast of tropical cyclone trajectories is crucial for the protection of people and property. Although forecast dynamical models can provide high-precision short-term forecasts, they are computationally demanding, and current statistical forecasting models have much room for improvement given that the database of past hurricanes is constantly growing. Machine learning methods, that can capture non-linearities and complex relations, have only been scarcely tested for this application. We propose a neural network model fusing past trajectory data and reanalysis atmospheric images (wind and pressure 3D fields). We use a moving frame of reference that follows the storm center for the 24h tracking forecast. The network is trained to estimate the longitude and latitude displacement of tropical cyclones and depressions from a large database from both hemispheres (more than 3000 storms since 1979, sampled at a 6 hour frequency). The advantage of the fused network is demonstrated and a comparison with current forecast models shows that deep learning methods could provide a valuable and complementary prediction. Moreover, our method can give a forecast for a new storm in a few seconds, which is an important asset for real-time forecasts compared to traditional forecasts.

Spurious samples in deep generative models: bug or feature?

Traditional wisdom in generative modeling literature is that spurious samples that a model can generate are errors and they should be avoided. Recent research, however, has shown interest in studying or even exploiting such samples instead of eliminating them. In this paper, we ask the question whether such samples can be eliminated all together without sacrificing coverage of the generating distribution. For the class of models we consider, we experimentally demonstrate that this is not possible without losing the ability to model some of the test samples. While our results need to be confirmed on a broader set of model families, these initial findings provide partial evidence that spurious samples share structural properties with the learned dataset, which, in turn, suggests they are not simply errors but a feature of deep generative nets.