Abstract To generate accurate and reliable predictions, modern AI systems need to combine data from multiple modalities, such as text, images, audio, spreadsheets, and time series. However, collecting training and test data for many modalities is challenging and time-consuming, creating a need for cost-efficient multi-modal data acquisition. In this paper we advocate that this can be realized by disentangling epistemic and aleatoric uncertainty. It is commonly assumed in the machine learning community that epistemic uncertainty can be reduced by collecting more data, while aleatoric uncertainty is irreducible. We claim that this assumption can be challenged in modern multi-modal AI systems, and we introduces an innovative data acquisition framework where uncertainty disentanglement leads to actionable decisions, allowing cost-efficient sampling in two directions: sample size and data modality. The main hypothesis is that aleatoric uncertainty decreases as the number of modalities increases, while epistemic uncertainty decreases by collecting more observations. We provide a theoretical analysis and proof-of-concept implementations on various multi-modal datasets to prove the usefulness of our framework, which combines ideas from active learning, active feature acquisition and uncertainty quantification.

Traditional machine learning models provide a single-class prediction for a given input instance. This may be inadequate in some scenarios, especially when the cost of erroneous predictions is high. Cautious random forests are cautious classification models that may output sets of possible classes as predictions when uncertainty is high, thus reducing the risk of making incorrect decisions. However, making such indeterminate predictions carries a cost, as resolving indeterminacy typically necessitates further analysis and manual intervention. This work focuses on explaining why an indeterminate prediction has been made and how indeterminacy can be resolved. To this end, we use counterfactual examples associated with determinate predictions. We propose a branch-and-bound algorithm that can efficiently generate proximal, plausible, and actionable counterfactual examples. Several experimental results are presented to demonstrate the advantages of our proposed method.

Probabilistic Circuits (PCs) are prominent tractable probabilistic models, allowing for a wide range of exact inferences. This paper focuses on a main algorithm for training PCs, LearnSPN, arguably a gold standard due to its efficiency, performance, and ease of use, in particular for tabular data. We show that LearnSPN is a greedy likelihood maximizer under mild assumptions. While inferences in PCs may use the entire circuit structure for processing queries, LearnSPN applies a hard method for learning PCs, propagating at each sum node a data point through one and only one of the children/edges as in a hard clustering process. We propose a new learning procedure named SoftLearn, that induces a PC using a soft clustering process. We investigate the effect of this learning-inference compatibility in PCs. Our experiments show that SoftLearn outperforms LearnSPN in many situations, yielding better likelihoods and arguably better samples. We also analyze comparable tractable models to highlight the differences between soft/hard learning and model querying.

International audience

The Plackett–Luce model is a popular parametric probabilistic model to define distributions between rankings of objects, modelling for instance observed preferences of users or ranked performances of algorithms. Since such observations may be scarce (users may provide partial preferences, or not all algorithms are run for a given experiment), it may be useful to consider the case where the parameters of the Plackett–Luce model are imprecisely known. In this paper, we first introduce the imprecise Plackett–Luce model, induced by a set of parameters (for instance, parameters with a high relative likelihood). Given a set of possible parameters for the model, we then provide an efficient algorithm to make cautious inferences, returning sets of possible optimal rankings (for instance in the form of partial orders). We illustrate the use of our imprecise model on label ranking, a specific kind of supervised learning.

A plug-in hybrid electric vehicle (PHEV) satisfies the driver’s power demand with two types of energy potentials: fuel and electrical energy provided by a battery. Classically, the battery consumption is planned over a trip to minimize the expected fuel consumption. A cautious driver will save battery potential to cross restricted areas (with desired low or even zero fuel consumption) without the fuel engine. This paper proposes an approach to minimize energy consumption while controlling the risk of a PHEV falling short of battery potential when crossing a restricted area. We use a nonlinear Gaussian process, trained on real vehicle data, for predicting the vehicle consumption. We take into account prediction uncertainty by ensuring that the driver’s highest power demand will be satisfied with a high probability. The interest of the approach is demonstrated by a simulated trip around Paris.

This paper proposes to use Gaussian process regression to predict the consumption of a plug-in electric hybrid vehicle from low-quality data. We specify background knowledge regarding new operating points and information regarding the noise process. This makes it possible to adapt the original (naive’) model. Experiments realized using dynamic and energetic models simulated electrified vehicle show the interest of our approach in order to improve robustness against scarce and noisy data.

This paper presents a novel approach for predicting the energy consumption of a plug-in hybrid electric vehicle (PHEV). We propose to estimate energy consumption strategy from data via regression applied to trip recordings. Descriptors of the trip elements are obtained from both recordings and statistics provided by a GPS navigation system. Trips are then split into elementary units corresponding to an homogeneous driving context. For each trip element, the optimal energy consumption strategy is computed via (expensive) dynamic programming simulations. Here, data analysis is used so as to identify descriptors of this trip element that are relevant to predict the energy consumption. Then, a polynomial model is fit to the data so as to estimate, for each new trip element, the optimal energy consumption strategy from the expected driving condition, rather than using dynamic programming. Our approach distinguishes itself by the fact that road context, driver style, road slope and auxiliary electrical power are taken into account to estimate the energy consumption of a PHEV. The accuracy of the prediction process is evaluated over test data, and demonstrates the interest of our approach in predicting energy consumption.

Supervised classification algorithms such as Boosting and SVM have achieved significant success in the field of computer vision for classification and object recognition. However, the performance of the classifier decreases rapidly if there are insufficient labeled training samples. In this paper, a semi-supervised boosting algorithm is proposed to overcome this limitation. First, a few labeled instances are use to estimate probabilistic class labels for unlabeled samples using Gaussian Mixture Models after a dimension reduction step performed via Principal Component Analysis. Then, we apply a boosting strategy on decision stumps trained using the soft labeled instances thus obtained. The performances of our strategy are evaluated on several state-of-the-art classification datasets, as well as on a pedestrian detection and recognition problem. Experimental results demonstrate the interest of taking into account additional data in the training process.

In this paper, we propose an effective approach for moving object detection based on modeling the ego-motion uncertainty and using a graph-cut based motion segmentation. First, the relative camera pose is estimated by minimizing the sum of reprojection errors and its covariance matrix is calculated using a first-order errors propagation method. Next, a motion likelihood for each pixel is obtained by propagating the uncertainty of the ego-motion to the Residual Image Motion Flow (RIMF). Finally, the motion likelihood and the depth gradient are used in a graph-cut based approach as region and boundary terms respectively, in order to obtain the moving objects segmentation. Experimental results on real-world data show that our approach can detect dynamic objects which move on the epipolar plane or that are partially occluded in complex urban traffic scenes.

Fuzzy or hard partitioning methods aim at grouping objects according to their similarity. Recently, a new concept of partition based on belief function theory, called credal partition, has been proposed and has been shown to generate meaningful description of the data. Hard, fuzzy or credal partitions are generally obtained using unsupervised learning methods, using only the numeric description between two objects to compute their similarity. However, in some applications, some kind of background knowledge about the objects or about the clusters is available. To integrate this auxiliary information, constraint-based (or semi-supervised) methods have been proposed. A popular type of constraints specifies whether two objects are in the same cluster (must-link) or in different clusters (cannot-link). We propose here a new algorithm, called CECM, which computes a credal partition using a constrained clustering method. We show how to translate the available information into constraints, and how to integrate them in the search of the credal partition. The paper ends with some experimental results. Results of CECM are compared to other constrained clustering algorithms. Then an application in image segmentation is described.