A General Framework for Uncertainty Estimation in Deep Learning

arXiv.org Machine Learning

End-to-end learning has recently emerged as a promising technique to tackle the problem of autonomous driving. Existing works show that learning a navigation policy from raw sensor data may reduce the system's reliance on external sensing systems, (e.g. GPS), and/or outperform traditional methods based on state estimation and planning. However, existing end-to-end methods generally trade off performance for safety, hindering their diffusion to real-life applications. For example, when confronted with an input which is radically different from the training data, end-to-end autonomous driving systems are likely to fail, compromising the safety of the vehicle. To detect such failure cases, this work proposes a general framework for uncertainty estimation which enables a policy trained end-to-end to predict not only action commands, but also a confidence about its own predictions. In contrast to previous works, our framework can be applied to any existing neural network and task, without the need to change the network's architecture or loss, or to train the network. In order to do so, we generate confidence levels by forward propagation of input and model uncertainties using Bayesian inference. We test our framework on the task of steering angle regression for an autonomous car, and compare our approach to existing methods with both qualitative and quantitative results on a real dataset. Finally, we show an interesting by-product of our framework: robustness against adversarial attacks.

Perspectives on the Validation and Verification of Machine Learning Systems in the Context of Highly Automated Vehicles

AAAI Conferences

Algorithms incorporating learned functionality play an increasingly important role for highly automated vehicles. Their impressive performance within environmental perception and other tasks central to automated driving comes at the price of a hitherto unsolved functional verification problem within safety analysis. We propose to combine statistical guarantee statements about the generalisation ability of learning algorithms with the functional architecture as well as constraints about the dynamics and ontology of the physical world, yielding an integrated formulation of the safety verification problem of functional architectures comprising artificial intelligence components. Its formulation as a probabilistic constraint system enables calculation of low risk manoeuvres. We illustrate the proposed scheme on a simple automotive scenario featuring unreliable environmental perception.

Symbolic Traffic Scene Analysis Using Dynamic Belief Networks 1 Introduction

AAAI Conferences

An important task for progress in IVHS (Intelligent Vehicle Highway Systems) is the development of methods for real-time traffic scene analysis. All three major applications of IVHS - ADIS (Advanced Driver Information Systems), ATMS (Advanced Traffic Management Systems), and AVCS (Automated Vehicle Control Systems) - could benefit from accurate, high-level descriptions of traffic situations. For example, an ADIS and an ATMS could use information about traffic congestion and accidents to alert drivers or to direct vehicles to alternate routes. An ATMS also could analyze local traffic at intersections to identify those with higher risk of accidents. Finally, an AVCS would need information about the actions of neighboring vehicles and the condition of traffic lanes ahead to control an automated car moving along a freeway [6].

Closed-Loop Policies for Operational Tests of Safety-Critical Systems

arXiv.org Artificial Intelligence

Manufacturers of safety-critical systems must make the case that their product is sufficiently safe for public deployment. Much of this case often relies upon critical event outcomes from real-world testing, requiring manufacturers to be strategic about how they allocate testing resources in order to maximize their chances of demonstrating system safety. This work frames the partially observable and belief-dependent problem of test scheduling as a Markov decision process, which can be solved efficiently to yield closed-loop manufacturer testing policies. By solving for policies over a wide range of problem formulations, we are able to provide high-level guidance for manufacturers and regulators on issues relating to the testing of safety-critical systems. This guidance spans an array of topics, including circumstances under which manufacturers should continue testing despite observed incidents, when manufacturers should test aggressively, and when regulators should increase or reduce the real-world testing requirements for an autonomous vehicle.

Flexible Models for Microclustering with Application to Entity Resolution

Neural Information Processing Systems

Most generative models for clustering implicitly assume that the number of data points in each cluster grows linearly with the total number of data points. Finite mixture models, Dirichlet process mixture models, and Pitman-Yor process mixture models make this assumption, as do all other infinitely exchangeable clustering models. However, for some applications, this assumption is inappropriate. For example, when performing entity resolution, the size of each cluster should be unrelated to the size of the data set, and each cluster should contain a negligible fraction of the total number of data points. These applications require models that yield clusters whose sizes grow sublinearly with the size of the data set. We address this requirement by defining the microclustering property and introducing a new class of models that can exhibit this property. We compare models within this class to two commonly used clustering models using four entity-resolution data sets.