Corpus ID: 84186487

TATi-Thermodynamic Analytics ToolkIt: TensorFlow-based software for posterior sampling in machine learning applications

@article{Heber2019TATiThermodynamicAT,
  title={TATi-Thermodynamic Analytics ToolkIt: TensorFlow-based software for posterior sampling in machine learning applications},
  author={Frederik Heber and Zofia Trstanova and Benedict J. Leimkuhler},
  journal={ArXiv},
  year={2019},
  volume={abs/1903.08640}
}
  • Frederik Heber, Zofia Trstanova, Benedict J. Leimkuhler
  • Published 2019
  • Mathematics, Computer Science
  • ArXiv
  • We describe a TensorFlow-based library for posterior sampling and exploration in machine learning applications. TATi, the Thermodynamic Analytics ToolkIt, implements algorithms for 2nd order (underdamped) Langevin dynamics and Hamiltonian Monte Carlo (HMC). It also allows for rapid prototyping of new sampling methods in pure Python and supports an ensemble framework for generating multiple trajectories in parallel, a capability that is demonstrated by the implementation of a recently proposed… CONTINUE READING

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 90 REFERENCES

    TensorFlow Distributions

    Adaptive Thermostats for Noisy Gradient Systems

    VIEW 1 EXCERPT

    Intrinsic map dynamics exploration for uncharted effective free-energy landscapes

    Automatic differentiation in PyTorch

    VIEW 1 EXCERPT

    Optimization Methods for Large-Scale Machine Learning

    VIEW 1 EXCERPT

    Adam: A Method for Stochastic Optimization

    VIEW 2 EXCERPTS

    Stochastic Gradient Hamiltonian Monte Carlo

    VIEW 1 EXCERPT