Stochastic Gradient Descent Langevin Dynamics (SGLD) algorithms, which add noise to the classic gradient descent, are known to improve the training of neural networks in some cases where the neural network is very deep. In this paper we study the possibilities of training acceleration for the numerical resolution of stochastic control problems through gradient descent, where the control is parametrized by a neural network. If the control is applied at many discretization times then solving the stochastic control problem reduces to minimizing the loss of a very deep neural network. We numerically show that Langevin algorithms improve the training on various stochastic control problems like hedging and resource management, and for different choices of gradient descent methods.

We propose characteristics-informed neural networks (CINN), a simple and efficient machine learning approach for solving forward and inverse problems involving hyperbolic PDEs. Like physics-informed neural networks (PINN), CINN is a meshless machine learning solver with universal approximation capabilities. Unlike PINN, which enforces a PDE softly via a multi-part loss function, CINN encodes the characteristics of the PDE in a general-purpose deep neural network by adding a characteristic layer. This neural network is trained with the usual MSE data-fitting regression loss and does not require residual losses on collocation points. This leads to faster training and can avoid well-known pathologies of gradient descent optimization of multi-part PINN loss functions. This paper focuses on linear transport phenomena, in which case it is shown that, if the characteristic ODEs can be solved exactly, then the output of a CINN is an exact solution of the PDE, even at initialization, preventing the occurrence of non-physical solutions. In addition, a CINN can also be trained with soft penalty constraints that enforce, for example, periodic or Neumman boundary conditions, without losing the property that the output satisfies the PDE automatically. We also propose an architecture that extends the CINN approach to linear hyperbolic systems of PDEs. All CINN architectures proposed here can be trained end-to-end from sample data using standard deep learning software. Experiments with the simple advection equation, a stiff periodic advection equation, and an acoustics problem where data from one field is used to predict the other, unseen field, indicate that CINN is able to improve on the accuracy of the baseline PINN, in some cases by a considerable margin, while also being significantly faster to train and avoiding non-physical solutions. An extension to nonlinear PDEs is also briefly discussed.

Generating a Boltzmann distribution in high dimension has recently been achieved with Normalizing Flows, which enable fast and exact computation of the generated density, and thus unbiased estimation of expectations. However, current implementations rely on accurate training data, which typically comes from computationally expensive simulations. There is therefore a clear incentive to train models with incomplete or no data by relying solely on the target density, which can be obtained from a physical energy model (up to a constant factor). For that purpose, we analyze the properties of standard losses based on Kullback-Leibler divergences. We showcase their limitations, in particular a strong propensity for mode collapse during optimization on high-dimensional distributions. We then propose strategies to alleviate these issues, most importantly a new loss function well-grounded in theory and with suitable optimization properties. Using as a benchmark the generation of 3D molecular configurations, we show on several tasks that, for the first time, imperfect pre-trained models can be further optimized in the absence of training data.

Geomechanical monitoring of a rock massif is an actively developing branch of geomechanics. It is almost impossible to single out a methodology and approaches for data collection and analysis in developing seismic monitoring systems. In the process of mining in rock massif, changes in the state of structural inhomogeneities are most clearly manifested. Existing natural structural inhomogeneities are revealed, there are movements in discontinuous disturbances, and new technogenic disturbances are formed, which are accompanied by a change in the natural stress state of various blocks of the massif. An important task is to develop a mining forecasting model that can take into account the structural heterogeneity of the rock massif and select the necessary forecast horizon depending on monitoring data The developed method of evaluating the results of monitoring geomechanical processes in the rock massif allowed us to forecast of zones of possible rock bursts.

In many scientific research and engineering applications where repeated simulations of complex systems are conducted, a surrogate is commonly adopted to quickly estimate the whole system. To reduce the expensive cost of generating training examples, it has become a promising approach to combine the results of low-fidelity (fast but inaccurate) and high-fidelity (slow but accurate) simulations. Despite the fast developments of multi-fidelity fusion techniques, most existing methods require particular data structures and do not scale well to high-dimensional output. To resolve these issues, we generalize the classic autoregression (AR), which is wildly used due to its simplicity, robustness, accuracy, and tractability, and propose generalized autoregression (GAR) using tensor formulation and latent features. GAR can deal with arbitrary dimensional outputs and arbitrary multifidelity data structure to satisfy the demand of multi-fidelity fusion for complex problems; it admits a fully tractable likelihood and posterior requiring no approximate inference and scales well to high-dimensional problems. Furthermore, we prove the autokrigeability theorem based on GAR in the multi-fidelity case and develop CIGAR, a simplified GAR with the exact predictive mean accuracy with computation reduction by a factor of d 3, where d is the dimensionality of the output. The empirical assessment includes many canonical PDEs and real scientific examples and demonstrates that the proposed method consistently outperforms the SOTA methods with a large margin (up to 6x improvement in RMSE) with only a couple high-fidelity training samples.

This work introduces TRIGGER, the first lighTweight univeRsal jammInG Gripper for aErial gRasping. TRIGGER is an omnidirectional, landing-capable aerial grasping system with resilience and robustness to collisions and inherent passive compliance. In particular, this work presents the design, fabrication, and experimental validation of a novel, intelligent, modular, universal jamming gripper specifically designed for aerial applications. Leveraging recent developments in particle jamming and soft granular materials, TRIGGER produces 15N of holding force with only a relatively small activation force of 2.5N. Experiments show the relationship between fill ratio and activation force and reveal that adding an additive to the membrane's silicone mixture improves the holding force by up to 52%. The concept is validated by mounting TRIGGER onto a multicopter performing a pick-and-release task under laboratory conditions. Based on the experimental data, a model for robotic simulators is introduced to facilitate future controller developments.

Polynomial chaos expansion (PCE) is a versatile tool widely used in uncertainty quantification and machine learning, but its successful application depends strongly on the accuracy and reliability of the resulting PCE-based response surface. High accuracy typically requires high polynomial degrees, demanding many training points especially in high-dimensional problems through the curse of dimensionality. So-called sparse PCE concepts work with a much smaller selection of basis polynomials compared to conventional PCE approaches and can overcome the curse of dimensionality very efficiently, but have to pay specific attention to their strategies of choosing training points. Furthermore, the approximation error resembles an uncertainty that most existing PCE-based methods do not estimate. In this study, we develop and evaluate a fully Bayesian approach to establish the PCE representation via joint shrinkage priors and Markov chain Monte Carlo. The suggested Bayesian PCE model directly aims to solve the two challenges named above: achieving a sparse PCE representation and estimating uncertainty of the PCE itself. The embedded Bayesian regularizing via the joint shrinkage prior allows using higher polynomial degrees for given training points due to its ability to handle underdetermined situations, where the number of considered PCE coefficients could be much larger than the number of available training points. We also explore multiple variable selection methods to construct sparse PCE expansions based on the established Bayesian representations, while globally selecting the most meaningful orthonormal polynomials given the available training data. We demonstrate the advantages of our Bayesian PCE and the corresponding sparsity-inducing methods on several benchmarks.

As automated decision making and decision assistance systems become common in everyday life, research on the prevention or mitigation of potential harms that arise from decisions made by these systems has proliferated. However, various research communities have independently conceptualized these harms, envisioned potential applications, and proposed interventions. The result is a somewhat fractured landscape of literature focused generally on ensuring decision-making algorithms "do the right thing". In this paper, we compare and discuss work across two major subsets of this literature: algorithmic fairness, which focuses primarily on predictive systems, and ethical decision making, which focuses primarily on sequential decision making and planning. We explore how each of these settings has articulated its normative concerns, the viability of different techniques for these different settings, and how ideas from each setting may have utility for the other.

The flexibility introduced with the Open Radio Access Network (O-RAN) architecture allows us to think beyond static configurations in all parts of the network. This paper addresses the issue related to predicting the power consumption of different radio schedulers, and the potential offered by O-RAN to collect data, train models, and deploy policies to control the power consumption. We propose a black-box (Neural Network) model to learn the power consumption function. We compare our approach with a known hand-crafted solution based on domain knowledge. Our solution reaches similar performance without any previous knowledge of the application and provides more flexibility in scenarios where the system behavior is not well understood or the domain knowledge is not available.

Very large state spaces with a sparse reward signal are difficult to explore. The lack of a sophisticated guidance results in a poor performance for numerous reinforcement learning algorithms. In these cases, the commonly used random exploration is often not helpful. The literature shows that this kind of environments require enormous efforts to systematically explore large chunks of the state space. Learned state representations can help here to improve the search by providing semantic context and build a structure on top of the raw observations. In this work we introduce a novel time-myopic state representation that clusters temporal close states together while providing a time prediction capability between them. By adapting this model to the Go-Explore paradigm (Ecoffet et al., 2021b), we demonstrate the first learned state representation that reliably estimates novelty instead of using the hand-crafted representation heuristic. Our method shows an improved solution for the detachment problem which still remains an issue at the Go-Explore Exploration Phase. We provide evidence that our proposed method covers the entire state space with respect to all possible time trajectories without causing disadvantageous conflict-overlaps in the cell archive. Analogous to native Go-Explore, our approach is evaluated on the hard exploration environments MontezumaRevenge, Gravitar and Frostbite (Atari) in order to validate its capabilities on difficult tasks. Our experiments show that time-myopic Go-Explore is an effective alternative for the domain-engineered heuristic while also being more general. The source code of the method is available on GitHub.