Scaling re-done images tutorials

docs/tutorials/AE_MNIST.rst

@@ -27,30 +27,35 @@ The code_ given below produces the following output that is quite similar to the
 Visualization of 100 test samples.
 
 .. figure:: images/AE_MNIST_DATA.png
+   :scale: 75 %
    :align: center
    :alt: 100 MNIST digits (test data)
 
 The learned filters without regularization.
 
 .. figure:: images/AE_MNIST_FILTER.png
+   :scale: 75 %
    :align: center
    :alt: AE filter on MNIST
 
 The corresponding reconstruction of the model, that is the encoding followed by the decoding.
 
 .. figure:: images/AE_MNIST_REC.png
+   :scale: 75 %
    :align: center
    :alt: AE filter on MNIST
 
 The learned filters when a contractive penalty is used, leading to much more localized and less noisy filters.
 
 .. figure:: images/AE_MNIST_FILTER_CON.png
+   :scale: 75 %
    :align: center
    :alt: AE filter on MNIST with contrastive penalty
 
 And the corresponding reconstruction of the model.
 
 .. figure:: images/AE_MNIST_REC_CON.png
+   :scale: 75 %
    :align: center
    :alt: AE filter on MNIST with contrastive penalty
 

docs/tutorials/AE_natural_images.rst

@@ -24,49 +24,57 @@ The code_ given below produces the following output that is impressively similar
 Visualization of 100 examples of the gray scale natural image dataset.
 
 .. figure:: images/SAE_natural_images_data_raw.png
+   :scale: 75 %
    :align: center
    :alt: 100 gray scale natural image patch examples
 
 The corresponding whitened image patches.
 
 .. figure:: images/SAE_natural_images_data.png
+   :scale: 75 %
    :align: center
    :alt: 100 gray scale natural image patch examples whitened
 
 The learned filters from the whitened natural image patches.
 
 .. figure:: images/SAE_natural_images_filter.png
+   :scale: 75 %
    :align: center
    :alt: ICA filter on natural images
 
 The corresponding reconstruction of the model, that is the encoding followed by the decoding.
 
 .. figure:: images/SAE_natural_images_reconstruction.png
+   :scale: 75 %
    :align: center
    :alt: ICA filter on natural images
 
 To analyze the optimal response of the learn filters we can fit a Gabor-wavelet parametrized in angle and frequency, and plot
 the optimal grating, here for 20 filters
 
 .. figure:: images/SAE_natural_images_grating.png
+   :scale: 75 %
    :align: center
    :alt: ICA filters with fitted Gabor-wavelets.
 
 as well as the corresponding tuning curves, which show the responds/activities as a function frequency in pixels/cycle (left) and angle in rad (right).
 
 .. figure:: images/SAE_natural_images_tuning_curves.png
+   :scale: 75 %
    :alt: ICA  fiter's tuning curves
    :align: center
 
 Furthermore, we can plot the histogram of all filters over the frequencies in pixels/cycle (left) and angles in rad (right).
 
 .. figure:: images/SAE_natural_images_histogram.png
+   :scale: 75 %
    :alt: ICA histogram of frequency and angle
    :align: center
 
 We can also train the model on the unwhitened data leading to the following filters that cover also lower frequencies.
 
 .. figure:: images/SAE_natural_images_filter_unwhitened.png
+   :scale: 75 %
    :alt: ICA histogram of frequency and angle
    :align: center
 

docs/tutorials/GRBM_natural_images.rst

@@ -21,18 +21,21 @@ The code_ given below produces the following output.
 Visualization of the learned filters, which are very similar to those of ICA.
 
 .. figure:: images/GRBM_weights_unnormalized.png
+   :scale: 50 %
    :align: center
    :alt: GRBM weights unnormalized
 
 For a better visualization of the structure, here are the same filters normalized independently.
 
 .. figure:: images/GRBM_weights_normalized.png
+   :scale: 50 %
    :align: center
    :alt: GRBM weights normalized
 
 Sampling results for some examples. The first row shows some training data and the following rows are the results after one step of Gibbs-sampling starting from the previous row.
 
 .. figure:: images/GRBM_samples.png
+   :scale: 35 %
    :alt: GRBM samples
    :align: center
 
@@ -48,18 +51,21 @@ To analyze the optimal response of the learn filters we can fit a Gabor-wavelet
 the optimal grating, here for 20 filters
 
 .. figure:: images/GRBM_grating.png
+   :scale: 75 %
    :align: center
    :alt: GRBM filters with fitted Gabor-wavelets.
 
 as well as the corresponding tuning curves, which show the responds/activities as a function frequency in pixels/cycle (left) and angle in rad (right).
 
 .. figure:: images/GRBM_tuning.png
+   :scale: 75 %
    :alt: GRBM  fiter's tuning curves
    :align: center
 
 Furthermore, we can plot the histogram of all filters over the frequencies in pixels/cycle (left) and angles in rad (right).
 
 .. figure:: images/GRBM_histogram.png
+   :scale: 75 %
    :align: center
    :alt: GRBM histogram of frequency and angle
 

docs/tutorials/ICA_natural_images.rst

@@ -21,18 +21,21 @@ The code_ given below produces the following output.
 Visualization of 100 examples of the gray scale natural image dataset.
 
 .. figure:: images/ICA_natural_images_data.png
+   :scale: 35 %
    :align: center
    :alt: 100 gray scale natural image patch examples
 
 The corresponding whitened image patches.
 
 .. figure:: images/ICA_natural_images_data_whitened.png
+   :scale: 35 %
    :align: center
    :alt: 100 gray scale natural image patch examples whitened
 
 The learned filters/independent components learned from the whitened natural image patches.
 
 .. figure:: images/ICA_natural_images_filter.png
+   :scale: 35 %
    :align: center
    :alt: ICA filter on natural images
 
@@ -46,18 +49,21 @@ To analyze the optimal response of the learn filters we can fit a Gabor-wavelet
 the optimal grating, here for 20 filters
 
 .. figure:: images/ICA_grating.png
+   :scale: 35 %
    :align: center
    :alt: ICA filters with fitted Gabor-wavelets.
 
 as well as the corresponding tuning curves, which show the responds/activities as a function frequency in pixels/cycle (left) and angle in rad (right).
 
 .. figure:: images/ICA_tuning.png
+   :scale: 35 %
    :alt: ICA  fiter's tuning curves
    :align: center
 
 Furthermore, we can plot the histogram of all filters over the frequencies in pixels/cycle (left) and angles in rad (right).
 
 .. figure:: images/ICA_histogram.png
+   :scale: 35 %
    :alt: ICA histogram of frequency and angle
    :align: center
 

docs/tutorials/PCA_eigenfaces.rst

@@ -18,12 +18,14 @@ The code_ given below produces the following output.
 Some examples of the face images of the olivetti face dataset.
 
 .. figure:: images/example_faces.png
+   :scale: 75 %
    :align: center
    :alt: Examples of the face datset
 
 The first 100 principal components extracted from the dataset. The components focus on characteristics like glasses, lighting direction, nose shape, ...
 
 .. image:: images/components_faces.png
+   :scale: 75 %
    :align: center
    :alt: Principal components of teh face dataset
 
@@ -37,12 +39,14 @@ The cumulative sum of the Eigenvalues show how 'compressable' the dataset is.
 For example using only the first 50 eigenvectors retains 87,5 % of the variance of data and the reconstructed images look as follows.
 
 .. image:: images/reconstruction50.png
+   :scale: 75 %
    :align: center
    :alt: Reconstruction using 50 PCs
 
 For 200 eigenvectors we retain 98,0 % of the variance of the data and the reconstructed images look as follows.
 
 .. image:: images/reconstruction50.png
+   :scale: 75 %
    :align: center
    :alt: Reconstruction using 200 PCs
 

docs/tutorials/RBM_MNIST_big.rst

@@ -23,14 +23,14 @@ Learned filters of a centered binary RBM with 500 hidden units on the MNIST data
 The filters have been normalized such that the structure is more prominent.
 
 .. figure:: images/BRBM_big_centered_weights.png
-   :scale: 30 %
+   :scale: 35 %
    :alt: weights centered
    :align: center
 
 Sampling results for some examples. The first row shows some training data and the following rows are the results after one Gibbs-sampling step starting from the previous row.
 
 .. figure:: images/BRBM_big_centered_samples.png
-   :scale: 30 %
+   :scale: 35 %
    :alt: samples centered
    :align: center
 
@@ -46,14 +46,14 @@ Now we have a look at the filters learned for a normal binary RBM with 500 hidde
 The filters have also been normalized such that the structure is more prominent.
 
 .. figure:: images/BRBM_big_normal_weights.png
-   :scale: 30 %
+   :scale: 35 %
    :alt: weights centered
    :align: center
 
 Sampling results for some examples. The first row shows the training data and the following rows are the results after one Gibbs-sampling step starting from the previous row.
 
 .. figure:: images/BRBM_big_normal_samples.png
-   :scale: 30 %
+   :scale: 35%
    :alt: samples centered
    :align: center
 

docs/tutorials/RBM_MNIST_small.rst

@@ -45,6 +45,7 @@ The filters have been normalized such that the structure is more prominent.
 Sampling results for some examples. The first row shows the training data and the following rows are the results after one Gibbs-sampling step starting from the previous row.
 
 .. figure:: images/BRBM_small_centered_samples.png
+   :scale: 35 %
    :alt: samples centered
    :align: center
 
@@ -70,6 +71,7 @@ Resulting in the following weights and sampling steps.
    :align: center
 
 .. figure:: images/BRBM_small_normal_samples.png
+   :scale: 35 %
    :alt: samples normal
    :align: center
 
@@ -101,6 +103,7 @@ While the centered model has a similar performance on the flipped version,
    :align: center
 
 .. figure:: images/BRBM_small_centered_samples_flipped.png
+   :scale: 35 %
    :alt: flipped samples centered
    :align: center
 
@@ -118,6 +121,7 @@ The normal RBM has not.
    :align: center
 
 .. figure:: images/BRBM_small_normal_samples_flipped.png
+   :scale: 35 %
    :alt: flipped samples normal
    :align: center