We have three versions for the Santa algorithm: a MATLAB version for feedforward neural networks and convolutional neural networks, a Python version for recurrent neural networks, and a Caffe version for scalable learning implemented in C++. Please refer to the code for details.
This is a MATLAB implementation for the symmetric splitting integrators to improve convergence rates of SG-MCMC algorithms.
This is a MATLAB implementation for Scalable Bayesian Non-Negative Tensor Factorization.
This is a MATLAB implementation for Scalable Deep Poisson Factor Analysis.
This is a C++ implementation for stochastic gradient thermostats, applied for latent Dirichlet allocation (LDA).
This is a C++ implementation for Bayesian max-margin clustering, applied for cluster topic models.
This is a C++ implementation for differential topic models.
This is a MATLAB implementation with C++ mex to accelerate sampling for dependent normalized random measures.