Releases: alan-turing-institute/ThermodynamicAnalyticsToolkit
Releases · alan-turing-institute/ThermodynamicAnalyticsToolkit
Version 0.2
ChangeLog update
- enhanced userguide with fully tested python examples
- using double floating point precision by default
- allowed external setting of inter and intra ops thread number for
parallelizaton - distributable tarball is working (make dist), target distcheck not yet
- added measuring of execution times (init, train, overall)
- FIX: batch_size went missing
- FIX: Dataset was not shuffled properly and split into test/train inconsistent
- added loss manifold sampler
- added Python interface
- added userguide
- added saving and restoring of neural network model to and from file.
- added trajectory analyser for average parameters and diffusion map analysis
based on the contributed code by Zofia Trstanova - added sampling of average kinetic energy and configurational temperature,
accurate by accumulating in every step (not just "every_nth" step) - added autotools testsuite
- added SGLD, Geometric Langevin Algorithm 1st and 2nd order sampling
sha1: c1a663adb4ead4e5989cff57c16e3bf530b0ecc4 datadrivensampler-0.3.tar.bz2
Version 0.1
ChangeLog update
- gradient descent, SGD, and SGLD optimization
- datasets confined to ones from TensorFlow "playground"
- trajectory analyser