Link to default dev branch

Author: krasserm

README.md

@@ -5,42 +5,42 @@
 This repository is a collection of notebooks related to *Bayesian Machine Learning*. The following links display 
 the notebooks via [nbviewer](https://nbviewer.jupyter.org/) to ensure a proper rendering of formulas.
 
-- [Variational inference in Bayesian neural networks](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/bayesian-neural-networks/bayesian_neural_networks.ipynb). 
+- [Variational inference in Bayesian neural networks](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/bayesian-neural-networks/bayesian_neural_networks.ipynb). 
   Demonstrates how to implement a Bayesian neural network and variational inference of network parameters. Example implementation 
   with Keras (see also 
-  [PyMC4 implementation](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/bayesian-neural-networks/bayesian_neural_networks_pymc4.ipynb)).
+  [PyMC4 implementation](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/bayesian-neural-networks/bayesian_neural_networks_pymc4.ipynb)).
 
-- [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/krasserm/bayesian-machine-learning/blob/master/latent-variable-models/latent_variable_models_part_2.ipynb)
-  [Latent variable models, part 2: Stochastic variational inference and variational autoencoders](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/latent-variable-models/latent_variable_models_part_2.ipynb). 
+- [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/krasserm/bayesian-machine-learning/blob/dev/latent-variable-models/latent_variable_models_part_2.ipynb)
+  [Latent variable models, part 2: Stochastic variational inference and variational autoencoders](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/latent-variable-models/latent_variable_models_part_2.ipynb). 
   Introduction to stochastic variational inference with variational autoencoder as application example. Implementation 
   with Tensorflow 2.x.
 
-- [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/krasserm/bayesian-machine-learning/blob/master/latent-variable-models/latent_variable_models_part_1.ipynb)
-  [Latent variable models, part 1: Gaussian mixture models and the EM algorithm](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/latent-variable-models/latent_variable_models_part_1.ipynb).
+- [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/krasserm/bayesian-machine-learning/blob/dev/latent-variable-models/latent_variable_models_part_1.ipynb)
+  [Latent variable models, part 1: Gaussian mixture models and the EM algorithm](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/latent-variable-models/latent_variable_models_part_1.ipynb).
   Introduction to the expectation maximization (EM) algorithm and its application to Gaussian mixture models. Example
   implementation with plain NumPy/SciPy and scikit-learn for comparison (see also 
-  [PyMC3 implementation](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/latent-variable-models/latent_variable_models_part_1_pymc3.ipynb)).
+  [PyMC3 implementation](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/latent-variable-models/latent_variable_models_part_1_pymc3.ipynb)).
 
-- [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/krasserm/bayesian-machine-learning/blob/master/bayesian-optimization/bayesian_optimization.ipynb)
-  [Bayesian optimization](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/bayesian-optimization/bayesian_optimization.ipynb). 
+- [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/krasserm/bayesian-machine-learning/blob/dev/bayesian-optimization/bayesian_optimization.ipynb)
+  [Bayesian optimization](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/bayesian-optimization/bayesian_optimization.ipynb). 
   Introduction to Bayesian optimization. Example implementations with plain NumPy/SciPy as well as with libraries 
   scikit-optimize and GPyOpt. Hyper-parameter tuning as application example.  
 
-- [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/krasserm/bayesian-machine-learning/blob/master/gaussian-processes/gaussian_processes.ipynb)
-  [Gaussian processes](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/gaussian-processes/gaussian_processes.ipynb). 
+- [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/krasserm/bayesian-machine-learning/blob/dev/gaussian-processes/gaussian_processes.ipynb)
+  [Gaussian processes](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/gaussian-processes/gaussian_processes.ipynb). 
   Introduction to Gaussian processes. Example implementations with plain NumPy/SciPy as well as with libraries 
   scikit-learn and GPy. 
 
-- [Bayesian regression with linear basis function models](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/bayesian-linear-regression/bayesian_linear_regression.ipynb). 
+- [Bayesian regression with linear basis function models](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/bayesian-linear-regression/bayesian_linear_regression.ipynb). 
   Introduction to Bayesian linear regression. Implementation from scratch with plain NumPy as well as usage of scikit-learn 
   for comparison (see also 
   [PyMC4 implementation](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/maste/bayesian-linear-regressionr/bayesian_linear_regression_pymc4.ipynb) and 
-  [PyMC3 implementation](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/bayesian-linear-regression/bayesian_linear_regression_pymc3.ipynb)).
+  [PyMC3 implementation](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/bayesian-linear-regression/bayesian_linear_regression_pymc3.ipynb)).
 
-- [Deep feature consistent variational autoencoder](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder_dfc.ipynb). 
+- [Deep feature consistent variational autoencoder](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder_dfc.ipynb). 
   Describes how a perceptual loss can improve the quality of images generated by a variational autoencoder. Example 
   implementation with Keras.  
 
-- [Conditional generation via Bayesian optimization in latent space](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder_opt.ipynb). 
+- [Conditional generation via Bayesian optimization in latent space](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder_opt.ipynb). 
   Describes an approach for conditionally generating outputs with desired properties by doing Bayesian optimization in 
   latent space learned by a variational autoencoder. Example application implemented with Keras and GPyOpt.

autoencoder-applications/variational_autoencoder.ipynb

@@ -8,10 +8,10 @@
     "\n",
     "**Update, Dec. 17<sup>th</sup> 2019**: This notebook is superseded by the following two notebooks:\n",
     "\n",
-    "- [Latent variable models - part 1: Gaussian mixture models and the EM algorithm](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/latent-variable-models/latent_variable_models_part_1.ipynb)\n",
-    "- [Latent variable models - part 2: Stochastic variational inference and variational autoencoders](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/latent-variable-models/latent_variable_models_part_2.ipynb).\n",
+    "- [Latent variable models - part 1: Gaussian mixture models and the EM algorithm](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/latent-variable-models/latent_variable_models_part_1.ipynb)\n",
+    "- [Latent variable models - part 2: Stochastic variational inference and variational autoencoders](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/latent-variable-models/latent_variable_models_part_2.ipynb).\n",
     "\n",
-    "The following old variational autoencoder code<sup>[1]</sup> is still used in [other](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder_opt.ipynb) [notebooks](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder_dfc.ipynb) and kept here for further reference."
+    "The following old variational autoencoder code<sup>[1]</sup> is still used in [other](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder_opt.ipynb) [notebooks](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder_dfc.ipynb) and kept here for further reference."
    ]
   },
   {

autoencoder-applications/variational_autoencoder_dfc.ipynb

@@ -16,7 +16,7 @@
     "\n",
     "### Plain VAE\n",
     "\n",
-    "In a [previous article](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master//autoencoder-applications/variational_autoencoder.ipynb) I introduced the variational autoencoder (VAE) and how it can be trained with a variational lower bound $\\mathcal{L}$ as optimization objective using stochastic gradient ascent methods. In context of stochastic gradient descent its negative value is used as loss function $L_{vae}$ which is a sum of a reconstruction loss $L_{rec}$ and a regularization term $L_{kl}$:\n",
+    "In a [previous article](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev//autoencoder-applications/variational_autoencoder.ipynb) I introduced the variational autoencoder (VAE) and how it can be trained with a variational lower bound $\\mathcal{L}$ as optimization objective using stochastic gradient ascent methods. In context of stochastic gradient descent its negative value is used as loss function $L_{vae}$ which is a sum of a reconstruction loss $L_{rec}$ and a regularization term $L_{kl}$:\n",
     "\n",
     "$$\n",
     "\\begin{align*}\n",
@@ -87,7 +87,7 @@
    "source": [
     "## Training\n",
     "\n",
-    "In contrast to the original paper we will use the MNIST handwritten digits dataset for training and for demonstrating how a perceptual loss improves over a pixel-by-pixel reconstruction loss. We can therefore reuse the VAE [encoder](https://github.com/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder_opt_util.py#L15-L36) and [decoder](https://github.com/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder_opt_util.py#L38-L53) architectures from the already mentioned [previous article](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder.ipynb). The perceptual model is a [small CNN](https://github.com/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder_opt_util.py#L91-L105) (Fig. 3) that has already been trained in [another context](http://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder_opt.ipynb#Optimization-objectives) to classify MNIST images."
+    "In contrast to the original paper we will use the MNIST handwritten digits dataset for training and for demonstrating how a perceptual loss improves over a pixel-by-pixel reconstruction loss. We can therefore reuse the VAE [encoder](https://github.com/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder_opt_util.py#L15-L36) and [decoder](https://github.com/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder_opt_util.py#L38-L53) architectures from the already mentioned [previous article](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder.ipynb). The perceptual model is a [small CNN](https://github.com/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder_opt_util.py#L91-L105) (Fig. 3) that has already been trained in [another context](http://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder_opt.ipynb#Optimization-objectives) to classify MNIST images."
    ]
   },
   {

autoencoder-applications/variational_autoencoder_opt.ipynb

@@ -29,7 +29,7 @@
     "- How can application of Bayesian optimization methods be justified?\n",
     "- What are possible alternatives to this approach? \n",
     "\n",
-    "I'll leave experiments with the chemical compounds dataset and the public chemical VAE for another article. The following assumes some basic familiarity with [variational autoencoders](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder.ipynb), [Bayesian otpimization](http://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/bayesian-optimization/bayesian_optimization.ipynb) and [Gaussian processes](http://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/gaussian-processes/gaussian_processes.ipynb). For more information on these topics you may want to read the linked articles."
+    "I'll leave experiments with the chemical compounds dataset and the public chemical VAE for another article. The following assumes some basic familiarity with [variational autoencoders](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder.ipynb), [Bayesian otpimization](http://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/bayesian-optimization/bayesian_optimization.ipynb) and [Gaussian processes](http://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/gaussian-processes/gaussian_processes.ipynb). For more information on these topics you may want to read the linked articles."
    ]
   },
   {
@@ -44,13 +44,13 @@
     "\n",
     "### Encoder\n",
     "\n",
-    "The encoder is a CNN, identical to the one presented in the the [variational autoencoder](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder.ipynb) notebook.\n",
+    "The encoder is a CNN, identical to the one presented in the the [variational autoencoder](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder.ipynb) notebook.\n",
     "\n",
     "![encoder](images/vae-opt/encoder.png) \n",
     "\n",
     "### Decoder\n",
     "\n",
-    "The decoder is a CNN, identical to the one presented in the the [variational autoencoder](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder.ipynb) notebook.\n",
+    "The decoder is a CNN, identical to the one presented in the the [variational autoencoder](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder.ipynb) notebook.\n",
     "\n",
     "![decoder](images/vae-opt/decoder.png)\n",
     "\n",
@@ -90,7 +90,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Code for the encoder and decoder have already been presented [elsewhere](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/autoencoder-applications/variational_autoencoder.ipynb), so only code for the predictor is shown here (see [variational_autoencoder_opt_util.py](variational_autoencoder_opt_util.py) for other function definitions):"
+    "Code for the encoder and decoder have already been presented [elsewhere](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/autoencoder-applications/variational_autoencoder.ipynb), so only code for the predictor is shown here (see [variational_autoencoder_opt_util.py](variational_autoencoder_opt_util.py) for other function definitions):"
    ]
   },
   {

bayesian-linear-regression/bayesian_linear_regression_pymc3.ipynb

@@ -32,7 +32,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "This is a [PyMC3](https://docs.pymc.io/) implementation of the examples in [Bayesian regression with linear basis function models](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/bayesian-linear-regression/bayesian_linear_regression.ipynb). To recap, a linear regression model is a linear function of the parameters but not necessarily of the input. Input $x$ can be expanded with a set of non-linear basis functions $\\phi_j(x)$, where $(\\phi_1(x), \\dots, \\phi_M(x))^T = \\boldsymbol\\phi(x)$, for modeling a non-linear relationship between input $x$ and a function value $y$.\n",
+    "This is a [PyMC3](https://docs.pymc.io/) implementation of the examples in [Bayesian regression with linear basis function models](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/bayesian-linear-regression/bayesian_linear_regression.ipynb). To recap, a linear regression model is a linear function of the parameters but not necessarily of the input. Input $x$ can be expanded with a set of non-linear basis functions $\\phi_j(x)$, where $(\\phi_1(x), \\dots, \\phi_M(x))^T = \\boldsymbol\\phi(x)$, for modeling a non-linear relationship between input $x$ and a function value $y$.\n",
     "\n",
     "$$\n",
     "y(x, \\mathbf{w}) = w_0 + \\sum_{j=1}^{M}{w_j \\phi_j(x)} = w_0 + \\mathbf{w}_{1:}^T \\boldsymbol\\phi(x) \\tag{1}\n",

bayesian-linear-regression/bayesian_linear_regression_pymc4.ipynb

@@ -45,7 +45,7 @@
    "source": [
     "## Linear basis function models\n",
     "\n",
-    "The following is a PyMC4 implementation of [Bayesian regression with linear basis function models](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/master/bayesian-linear-regression/bayesian_linear_regression.ipynb). To recap, a linear regression model is a linear function of the parameters but not necessarily of the input. Input $x$ can be expanded with a set of non-linear basis functions $\\phi_j(x)$, where $(\\phi_1(x), \\dots, \\phi_M(x))^T = \\boldsymbol\\phi(x)$, for modeling a non-linear relationship between input $x$ and a function value $y$.\n",
+    "The following is a PyMC4 implementation of [Bayesian regression with linear basis function models](https://nbviewer.jupyter.org/github/krasserm/bayesian-machine-learning/blob/dev/bayesian-linear-regression/bayesian_linear_regression.ipynb). To recap, a linear regression model is a linear function of the parameters but not necessarily of the input. Input $x$ can be expanded with a set of non-linear basis functions $\\phi_j(x)$, where $(\\phi_1(x), \\dots, \\phi_M(x))^T = \\boldsymbol\\phi(x)$, for modeling a non-linear relationship between input $x$ and a function value $y$.\n",
     "\n",
     "$$\n",
     "y(x, \\mathbf{w}) = w_0 + \\sum_{j=1}^{M}{w_j \\phi_j(x)} = w_0 + \\mathbf{w}_{1:}^T \\boldsymbol\\phi(x) \\tag{1}\n",