Inferring Sparse, Overcomplete Image Codes Using an Efficient Coding Framework

By Michael S. Lewicki, Barbara Olshausen. 1998

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

End-to-end Optimized Image Compression (2016, background): “As in many other optimized representations – e.g., sparse coding (Lewicki and Olshausen, 1998) – as well as many engineered representations – e.g., the steerable pyramid (Simoncelli, Freeman, et al., 1992), curvelets (Cand ` es and Donoho, 2002), and dual-tree complex wavelets (Selesnick, Baraniuk,…”
Deep learning in neural networks: An overview (2014, background): “…(Falconbridge, Stamps, & Badcock, 2006; Földiák, 1990; Hinton & Ghahramani, 1997; Hochreiter & Schmidhuber, 1999; Hyvärinen, Hoyer, & Oja, 1999; Lewicki & Olshausen, 1998) throughwell-known feature detectors (e.g., Olshausen & Field, 1996; Schmidhuber, Eldracher, & Foltin, 1996), such as…”
A hierarchical neural system with attentional top-down enhancement of the spatial resolution for object recognition. (2000, background): “An overcomplete basis allows a greater number of basis vectors than inputs, and have been proposed because they achieve sparsity in the representation (Chen, Donoho & Saunders, 1996; Lee, 1996; Olshausen & Field, 1996; Lewicki & Olshausen, 1998 ).”
A Neurodynamical Model of Visual Attention: Feedback Enhancement of Spatial Resolution in a Hierarchical System (2001, background): “An overcomplete basis allows a greater number of basis vectors than inputs, which have been proposed because they achieve sparsity in the representation (Chen et al., 1996; Lewicki and Olshausen, 1998; Olshausen and Field, 1996; Lee, 1996).”
Blind source separation of more sources than mixtures using overcomplete representations (1999, background): “For example, Lewicki and Olshausen [10] used a two times overcomplete representation to find 2 144 basis vectors for 12 12 patches of natural images.”
Learning Overcomplete Representations (2000, methodology): “For the learning rule, we replaced the term A in equation A.39 with an approximation to λAATAH−1 (see equation A.36) as suggested in Lewicki and Olshausen (1998, 1999): −λATAH−1 ≈ I− BQ diag−1[V+QTBQ]QT, (7.1) where B = ∇s∇s log P(ŝ) and Q and V are obtained from the singular value decomposition…”
Independent component filters of natural images compared with simple cells in primary visual cortex (1998, background): “Current work on overcomplete bases (Olshausen & Field 1997; Lewicki & Olshausen 1998) may lead to similar variability when performed on a su¤ciently large and diverse image database.”