In this paper, we focus on exploiting neural networks for the analysis and planning stage in self-adaptive architectures. The studied motivating cases in the paper involve existing (legacy) self-adaptive architectures and their adaptation logic, which has been specified by logical rules. We further assume that there is a need to endow these systems with the ability to learn based on examples of inputs and expected outputs. One simple option to address such a need is to replace the reasoning based on logical rules with a neural network. However, this step brings several problems that often create at least a temporary regress. The reason is the logical rules typically represent a large and tested body of domain knowledge, which may be lost if the logical rules are replaced by a neural network. Further, the black-box nature of generic neural networks obfuscates how the systems work inside and consequently introduces more uncertainty. In this paper, we present a method that makes it possible to endow an existing self-adaptive architectures with the ability to learn using neural networks, while preserving domain knowledge existing in the logical rules. We introduce a continuum between the existing rule-based system and a system based on a generic neural network. We show how to navigate in this continuum and create a neural network architecture that naturally embeds the original logical rules and how to gradually scale the learning potential of the network, thus controlling the uncertainty inherent to all soft computing models. We showcase and evaluate the approach on representative excerpts from two larger real-life use cases.

Let the OSS Enterprise newsletter guide your open source journey! Meta (formerly Facebook) this week announced the release of Bean Machine, a probabilistic programming system that ostensibly makes it easier to represent and learn about uncertainties in AI models. Available in early beta, Bean Machine can be used to discover unobserved properties of a model via automatic, "uncertainty-aware" learning algorithms. "[Bean Machine is] inspired from a physical device for visualizing probability distributions, a pre-computing example of a probabilistic system," the Meta researchers behind Bean Machine explained in a blog post. "We on the Bean Machine development team believe that the usability of a system forms the bedrock for its success, and we've taken care to center Bean Machine's design around a declarative philosophy within the PyTorch ecosystem." It's commonly understood that deep learning models are overconfident -- even when they make mistakes.

Policy Search and Model Predictive Control~(MPC) are two different paradigms for robot control: policy search has the strength of automatically learning complex policies using experienced data, while MPC can offer optimal control performance using models and trajectory optimization. An open research question is how to leverage and combine the advantages of both approaches. In this work, we provide an answer by using policy search for automatically choosing high-level decision variables for MPC, which leads to a novel policy-search-for-model-predictive-control framework. Specifically, we formulate the MPC as a parameterized controller, where the hard-to-optimize decision variables are represented as high-level policies. Such a formulation allows optimizing policies in a self-supervised fashion. We validate this framework by focusing on a challenging problem in agile drone flight: flying a quadrotor through fast-moving gates. Experiments show that our controller achieves robust and real-time control performance in both simulation and the real world. The proposed framework offers a new perspective for merging learning and control.

Bayesian Networks are probabilistic graphical models that can compactly represent dependencies among random variables. Missing data and hidden variables require calculating the marginal probability distribution of a subset of the variables. While knowledge of the marginal probability distribution is crucial for various problems in statistics and machine learning, its exact computation is generally not feasible for categorical variables due to the NP-hardness of this task. We develop a divide-and-conquer approach using the graphical properties of Bayesian networks to split the computation of the marginal probability distribution into sub-calculations of lower dimensionality, reducing the overall computational complexity. Exploiting this property, we present an efficient and scalable algorithm for estimating the marginal probability distribution for categorical variables. The novel method is compared against state-of-the-art approximate inference methods in a benchmarking study, where it displays superior performance. As an immediate application, we demonstrate how the marginal probability distribution can be used to classify incomplete data against Bayesian networks and use this approach for identifying the cancer subtype of kidney cancer patient samples.

While learning the graphical structure of Bayesian networks from observational data is key to describing and helping understand data generating processes in complex applications, the task poses considerable challenges due to its computational complexity. The directed acyclic graph (DAG) representing a Bayesian network model is generally not identifiable from observational data, and a variety of methods exist to estimate its equivalence class instead. Under certain assumptions, the popular PC algorithm can consistently recover the correct equivalence class by testing for conditional independence (CI), starting from marginal independence relationships and progressively expanding the conditioning set. Here, we propose the dual PC algorithm, a novel scheme to carry out the CI tests within the PC algorithm by leveraging the inverse relationship between covariance and precision matrices. Notably, the elements of the precision matrix coincide with partial correlations for Gaussian data. Our algorithm then exploits block matrix inversions on the covariance and precision matrices to simultaneously perform tests on partial correlations of complementary (or dual) conditioning sets. The multiple CI tests of the dual PC algorithm, therefore, proceed by first considering marginal and full-order CI relationships and progressively moving to central-order ones. Simulation studies indicate that the dual PC algorithm outperforms the classical PC algorithm both in terms of run time and in recovering the underlying network structure.

Neural density estimators have proven remarkably powerful in performing efficient simulation-based Bayesian inference in various research domains. In particular, the BayesFlow framework uses a two-step approach to enable amortized parameter estimation in settings where the likelihood function is implicitly defined by a simulation program. But how faithful is such inference when simulations are poor representations of reality? In this paper, we conceptualize the types of model misspecification arising in simulation-based inference and systematically investigate the performance of the BayesFlow framework under these misspecifications. We propose an augmented optimization objective which imposes a probabilistic structure on the latent data space and utilize maximum mean discrepancy (MMD) to detect potentially catastrophic misspecifications during inference undermining the validity of the obtained results. We verify our detection criterion on a number of artificial and realistic misspecifications, ranging from toy conjugate models to complex models of decision making and disease outbreak dynamics applied to real data. Further, we show that posterior inference errors increase as a function of the distance between the true data-generating distribution and the typical set of simulations in the latent summary space. Thus, we demonstrate the dual utility of MMD as a method for detecting model misspecification and as a proxy for verifying the faithfulness of amortized Bayesian inference.

Machine learning (ML) methodology, fueled with access to ever-increasing masses of data and unprecedented computing power, has been the main driving factor of recent progress in artificial intelligence (AI) and its applications in various branches of science and technology, industry and business, economics and finance, amongst others. In this regard, ML is most commonly perceived as a means for predictive modeling, that is, for the data-driven construction of a model that is mainly used for the purpose of predicting unknown facts in a specific context -- albeit models may, of course, serve other purposes, too, such as understanding and explanation, or may have a more descriptive flavor. A predictive model, or "predictor" in ML jargon, is trained in a supervised manner on cases encountered by the "learner" over the course of time, such as emails categorized as spam or non-spam, and the model is then used to make predictions in future situations, e.g., to automatically mark new emails. Looking at emerging applications of ML methodology, there is a visible shift from predictive modeling to prescriptive modeling, by which we mean the task of learning a model that stipulates appropriate decisions or actions to be taken in real-world scenarios. In fact, decisions are nowadays increasingly automated and made by algorithms instead of humans, and most of these automated decision making (ADM) algorithms are trained on data using ML methods. For example, think of decisions in the context of employees recruitment, such as hiring or placement decisions [41].

In many application domains such as medicine, information retrieval, cybersecurity, social media, etc., datasets used for inducing classification models often have an unequal distribution of the instances of each class. This situation, known as imbalanced data classification, causes low predictive performance for the minority class examples. Thus, the prediction model is unreliable although the overall model accuracy can be acceptable. Oversampling and undersampling techniques are well-known strategies to deal with this problem by balancing the number of examples of each class. However, their effectiveness depends on several factors mainly related to data intrinsic characteristics, such as imbalance ratio, dataset size and dimensionality, overlapping between classes or borderline examples. In this work, the impact of these factors is analyzed through a comprehensive comparative study involving 40 datasets from different application areas. The objective is to obtain models for automatic selection of the best resampling strategy for any dataset based on its characteristics. These models allow us to check several factors simultaneously considering a wide range of values since they are induced from very varied datasets that cover a broad spectrum of conditions. This differs from most studies that focus on the individual analysis of the characteristics or cover a small range of values. In addition, the study encompasses both basic and advanced resampling strategies that are evaluated by means of eight different performance metrics, including new measures specifically designed for imbalanced data classification. The general nature of the proposal allows the choice of the most appropriate method regardless of the domain, avoiding the search for special purpose techniques that could be valid for the target data.

Probabilistic forecasting consists of stating a probability distribution for a future outcome based on past observations. In meteorology, ensembles of physics-based numerical models are run to get such distribution. Usually, performance is evaluated with scoring rules, functions of the forecast distribution and the observed outcome. With some scoring rules, calibration and sharpness of the forecast can be assessed at the same time. In deep learning, generative neural networks parametrize distributions on high-dimensional spaces and easily allow sampling by transforming draws from a latent variable. Conditional generative networks additionally constrain the distribution on an input variable. In this manuscript, we perform probabilistic forecasting with conditional generative networks trained to minimize scoring rule values. In contrast to Generative Adversarial Networks (GANs), no discriminator is required and training is stable. We perform experiments on two chaotic models and a global dataset of weather observations; results are satisfactory and better calibrated than what achieved by GANs.

Normalizing flows are diffeomorphic, typically dimension-preserving, models trained using the likelihood of the model. We use the SurVAE framework to construct dimension reducing surjective flows via a new layer, known as the funnel. We demonstrate its efficacy on a variety of datasets, and show it improves upon or matches the performance of existing flows while having a reduced latent space size. The funnel layer can be constructed from a wide range of transformations including restricted convolution and feed forward layers.