19 June 2017 Deep greedy learning under thermal variability in full diurnal cycles
Author Affiliations +
We study the generalization and scalability behavior of a deep belief network (DBN) applied to a challenging long-wave infrared hyperspectral dataset, consisting of radiance from several manmade and natural materials within a fixed site located 500 m from an observation tower. The collections cover multiple full diurnal cycles and include different atmospheric conditions. Using complementary priors, a DBN uses a greedy algorithm that can learn deep, directed belief networks one layer at a time and has two layers form to provide undirected associative memory. The greedy algorithm initializes a slower learning procedure, which fine-tunes the weights, using a contrastive version of the wake-sleep algorithm. After fine-tuning, a network with three hidden layers forms a very good generative model of the joint distribution of spectral data and their labels, despite significant data variability between and within classes due to environmental and temperature variation occurring within and between full diurnal cycles. We argue, however, that more questions than answers are raised regarding the generalization capacity of these deep nets through experiments aimed at investigating their training and augmented learning behavior.
© 2017 Society of Photo-Optical Instrumentation Engineers (SPIE)
Patrick Rauss, Patrick Rauss, Dalton Rosario, Dalton Rosario, } "Deep greedy learning under thermal variability in full diurnal cycles," Optical Engineering 56(8), 081809 (19 June 2017). https://doi.org/10.1117/1.OE.56.8.081809 . Submission: Received: 4 February 2017; Accepted: 31 May 2017
Received: 4 February 2017; Accepted: 31 May 2017; Published: 19 June 2017


Back to Top