论文信息 - LOCOCODE versus PCA and ICA

LOCOCODE versus PCA and ICA

We compare the performance of three unsupervised learning algorithms on visual patterns that are mixtures of few underlying sources: “Independent Component Analysis” (ICA), “Principal Component Analysis” (PCA), and our new method “Low-complexity coding and decoding” (LOCOCODE). ICA and PCA fail to separate the sources no matter whether their number is known or not. LOCOCODE, however, always separates them. It also codes with fewer bits per pixel than ICA and PCA.

Jürgen Schmidhuber | Sepp Hochreiter | S. Hochreiter | J. Schmidhuber

[1] Schuster,et al. Separation of a mixture of independent signals using time delayed correlations. , 1994, Physical review letters.

[2] R. Zemel,et al. Learning sparse multiple cause models , 2000, Proceedings 15th International Conference on Pattern Recognition. ICPR-2000.

[3] Geoffrey E. Hinton,et al. Generative models for discovering sparse distributed representations. , 1997, Philosophical transactions of the Royal Society of London. Series B, Biological sciences.

[4] Geoffrey E. Hinton,et al. The "wake-sleep" algorithm for unsupervised neural networks. , 1995, Science.

[5] Bruno A. Olshausen,et al. Inferring Sparse, Overcomplete Image Codes Using an Efficient Coding Framework , 1998, NIPS.

[6] Jürgen Schmidhuber,et al. Unsupervised Coding with LOCOCODE , 1997, ICANN.

[7] Jürgen Schmidhuber,et al. Feature Extraction Through LOCOCODE , 1999, Neural Computation.

[8] Terrence J. Sejnowski,et al. An Information-Maximization Approach to Blind Separation and Blind Deconvolution , 1995, Neural Computation.

[9] Andrzej Cichocki,et al. A New Learning Algorithm for Blind Signal Separation , 1995, NIPS.

[10] J. Cardoso,et al. Blind beamforming for non-gaussian signals , 1993 .

[11] Jürgen Schmidhuber,et al. Flat Minima , 1997, Neural Computation.

[12] David J. Field,et al. Emergence of simple-cell receptive field properties by learning a sparse code for natural images , 1996, Nature.