Output details
11 - Computer Science and Informatics
University College London
Bilinear modeling via augmented Lagrange multipliers (BALM).
<23>The open problem of constrained low-rank matrix factorization with missing entries arises in many different fields and
contexts: from photometric stereo to non-rigid 3d reconstruction, audio source separation or dictionary learning. Prior
solutions had to be tailored to the specific problem at hand. Our paper formulates and demonstrates the first approach that can handle seamlessly different constrained bilinear problems. An extended version of an ECCV’10 conference paper, it embodies a core result from my ERC STG-204871 Grant. Our publicly available code features in the ’Nuit Blanche’ online resource for reproducible research in Compressive Sensing and Advanced Matrix Factorization Techniques.