ThermodynamicAnalyticsToolkit
ThermodynamicAnalyticsToolkit copied to clipboard
alan-turing-institute
Sampling-based approach to analyse neural networks using TensorFlow
The sampler code in, e.g., src/TATi/samplers/baoabsampler.py adds nodes to the graph to perform a single sampling step for each of the parameter matrices. TATi creates two parameter matrices per layer,...
We had deactivated shuffling of the dataset prior to tf.Dataset introduction because of tf's shuffle-repeat "bug". Since tf.Dataset allows to first shuffle and then repeat, we may reactivate shuffling by...
At the moment the exploration code requires that the _parallel_replica_ parameter is set to 1. It would be possible to allow for multiple dependent walkers (e.g. for ensemble preconditioning within...
Metadata
Owner
Metadata
Sampling-based approach to analyse neural networks using TensorFlow