190 likes | 306 Views
Olshausen’s Demo. How Important Is:. The Training set ? Natural Images (Olhausen’s database) How much do we learn ? face database and car database The Sparseness term ? Prior steepness Sparseness function Natural encoding or hacking? Whitening the data Non-stationary hypothesis.
E N D
How Important Is: • The Training set ? • Natural Images (Olhausen’s database) • How much do we learn ? • face database and car database • The Sparseness term ? • Prior steepness • Sparseness function • Natural encoding or hacking? • Whitening the data • Non-stationary hypothesis
Training with Natural Images • Training: 10 images (512x512) • 10,000 presentations • Batch size: 100 • Basis Function: 16x16
Face Database • Training: 100 images (100x100) • 10,000 presentations • Batch size: 100 • Basis Function: 16x16
Encoding Properties Original 10 basis 20 basis 30 basis 40 basis 50 basis
Car Database • Training: 200 images (128x128) • 10,000 presentations • Batch size: 100 • Basis Function: 16x16
Comments • The algorithm seems to capture the structure of the images (cf car): • Learning is experience-dependent • Basis functions found in good agreement with properties of neurons in visual cortex: • Receptive fields are localized, oriented, bandpass
How Important Is: • The Training set ? • Background, face and car databases • The Sparseness term ? • Prior steepness • Sparseness function • Natural encoding or hacking? • Whitening the data • Non-stationary hypothesis
Prior Steepness Steepness 2.2 Steepness 5 Steepness 10 Steepness 100
Prior Steepness Steepness 2.2 Steepness 1.5 Steepness 0.2
Sparseness Function S(x)=log(1+x^2) S(x)=|x|
Sparseness Function • batch of 100 samples: • Mean Error: abs=.471 / log = .504
How Important Is: • The Training set ? • Background, face and car databases • The Sparseness term ? • Prior steepness • Sparseness function • Natural encoding or hacking? • Whitening the data • Non-stationary hypothesis
Whitening the Data Data are filtered with whitening/low-pass filter: • How important is it for the convergence of the algorithm? • The question is to know whether it is just a speed-up or is it required for convergence?
Non-preprocessed Car Images • Training: 100 images (100x100) • 30,000 presentations • Batch size: 100 • Basis Function: 16x16
Non-stationary Hypothesis:Encoding the Full Face After few iterations…