Description:

A bare-bones TensorFlow framework for Bayesian deep learning and Gaussian process approximation with stochastic gradient variational Bayes inference.

Some of the features of Aboleth:

• Bayesian fully-connected, embedding and convolutional layers using SGVB for inference.

• Random Fourier and arc-cosine features for approximate Gaussian processes. Optional variational optimisation of these feature weights.

• Imputation layers with parameters that are learned as part of a model.

• Very flexible construction of networks, e.g. multiple inputs, ResNets etc.

• Optional maximum-likelihood type II inference for model parameters such as weight priors/regularizers and regression observation noise.

• Compatible and interoperable with other neural net frameworks such as Keras (see the demos for more information).

Changes to previous version:

Release 0.7.0

• Update to TensorFlow r1.4.

• Tutorials in the documentation on:

• Interfacing with Keras

• Saving/loading models

• How to build a variety of regressors with Aboleth

• New prediction module with some convenience functions, including freezing the weight samples during prediction.

• Bayesian convolutional layers with accompanying demo.

• Allow the number of samples drawn from a model to be varied by using placeholders.

• Generalise the feature embedding layers to work on matrix inputs (instead of just column vectors).

• Numerous numerical and usability fixes.

BibTeX Entry: Download

Supported Operating Systems: Linux

Data Formats: Any

Tags: Deep Learning, Variational Inference, Gaussian Process, Tensorflow

Archive: download here

Other available revisons

Version	Changelog	Date
0.7	Release 0.7.0 Update to TensorFlow r1.4. Tutorials in the documentation on: Interfacing with Keras Saving/loading models How to build a variety of regressors with Aboleth New prediction module with some convenience functions, including freezing the weight samples during prediction. Bayesian convolutional layers with accompanying demo. Allow the number of samples drawn from a model to be varied by using placeholders. Generalise the feature embedding layers to work on matrix inputs (instead of just column vectors). Numerous numerical and usability fixes.	December 14, 2017, 02:39:19
0.6.2	Hotfix release fix random seeds fix dropout sampling layers	October 13, 2017, 01:21:35
0.6	Some moderate changes to the API from: Using TensorFlow's tf.distributions to replace Aboleth's likelihoods Using TensorFlow's tf.distributions to replace Aboleth's distributions	September 27, 2017, 10:12:09
0.5	Initial Announcement on mloss.org.	September 7, 2017, 03:57:39

Comments

No one has posted any comments yet. Perhaps you'd like to be the first?

Leave a comment

You must be logged in to post comments.