论文信息 - A maximum likelihood algorithm for slow features analysis

A maximum likelihood algorithm for slow features analysis

• Motivate temporal slowness as a useful learning principle• Introduce Slow Features Analysis (SFA) as a learning algorithm, based on thisprinciple and give some intuitions• We show how the SFA algorithm corresponds to maximum likelihood learningin a limit of a probabilistic model and illustrate with simulations• Suggest several extensions and improvements to the probabilistic modeland connect to other work.

Richard E. Turner | Maneesh Sahani

[1] Geoffrey E. Hinton,et al. Self-organizing neural network that discovers surfaces in random-dot stereograms , 1992, Nature.

[2] Sam T. Roweis,et al. EM Algorithms for PCA and SPCA , 1997, NIPS.

[3] Laurenz Wiskott. Estimating Driving Forces of Nonstationary Time Series with Slow Feature Analysis Laurenz Wiskott Institute for Theoretical Biology , 2003 .

[4] Laurenz Wiskott,et al. Slow feature analysis yields a rich repertoire of complex cell properties. , 2005, Journal of vision.

[5] Harri Valpola,et al. Denoising Source Separation , 2005, J. Mach. Learn. Res..

[6] Terrence J. Sejnowski,et al. Slow Feature Analysis: Unsupervised Learning of Invariances , 2002, Neural Computation.

[7] Laurenz Wiskott,et al. Applying Slow Feature Analysis to Image Sequences Yields a Rich Repertoire of Complex Cell Properties , 2002, ICANN.

[8] Konrad P. Körding,et al. Learning the Nonlinearity of Neurons from Natural Visual Stimuli , 2003, Neural Computation.

[9] Aapo Hyvärinen,et al. Simple-Cell-Like Receptive Fields Maximize Temporal Coherence in Natural Video , 2003, Neural Computation.