Planned intervention: On Wednesday April 3rd 05:30 UTC Zenodo will be unavailable for up to 2-10 minutes to perform a storage cluster upgrade.
Published April 30, 2020 | Version v1
Conference paper Open

INPUT COMPLEXITY AND OUT-OF-DISTRIBUTION DETECTION WITH LIKELIHOOD-BASED GENERATIVE MODELS

  • 1. Dolby Laboratories, Barcelona, Spain
  • 2. Universitat Politècnica de Catalunya, Barcelona, Spain
  • 3. Universitat Pompeu Fabra, Barcelona, Spain
  • 4. Telefónica I+D, Research, Barcelona, Spain

Description

Likelihood-based generative models are a promising resource to detect out-of- distribution (OOD) inputs which could compromise the robustness or reliability of a machine learning system. However, likelihoods derived from such models have been shown to be problematic for detecting certain types of inputs that sig- nificantly differ from training data. In this paper, we pose that this problem is due to the excessive influence that input complexity has in generative models’ likelihoods. We report a set of experiments supporting this hypothesis, and use an estimate of input complexity to derive an efficient and parameter-free OOD score, which can be seen as a likelihood-ratio, akin to Bayesian model compari- son. We find such score to perform comparably to, or even better than, existing OOD detection approaches under a wide range of data sets, models, model sizes, and complexity estimates.

Notes

Accepted for ICLR2020

Files

input_complexity_and_out_of_di.pdf

Files (1.9 MB)

Name Size Download all
md5:fa030c9fd1c34044adfdcce3c9cbcdb8
1.9 MB Preview Download

Additional details

Funding

ACCORDION – Adaptive edge/cloud compute and network continuum over a heterogeneous sparse edge infrastructure to support nextgen applications 871793
European Commission