课堂展示简约ppt模板.pptVIP

Learning the parts of objects by non-negative matrix factorization Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Contents 1 Three methods learn to represent a face 2 The feature of three different methods 3 NMF algorithm applied to semantic analysis 4 Dos and don’ts Professional English 5 Other uses of NMF Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 1 Three methods learn to represent a face Professional English NMF PCA VQ Non-negative matrix factorization Principal components analysis Vector quantization Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 2 The feature of three different methods Professional English NMF PCA VQ Its images are localized features that correspond better with intuitive notions of the parts of faces The basis images for PCA are ‘eigenfaces’, some of which resemble distorted versions of whole faces VQ discovers a basis consisting of prototypes, each of which is a whole face Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Figure 1 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Figure 2 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Figure 3 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Three methods in a matrix factorization framework The r columns of W are called basis images. Each column of H is called an encoding and is in one-to-one correspondence with a face in V. The differences between PCA, VQ and NMF arise from different constraints imposed on the matrix factors W and H Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.