Matthew D. Hoffman

Learn More
We develop an online variational Bayes (VB) algorithm for Latent Dirichlet Allocation (LDA). Online LDA is based on online stochastic optimization with a natural gradient step, which we show converges to a local optimum of the VB objective function. It can handily analyze massive document collections, including those arriving in a stream. We study the(More)
The distinction between local and global variables will be important for us to develop online inference. In Bayesian statistics, for example, think of β as parameters with a prior and z1:n as hidden variables which are individual to each observation. In a Bayesian mixture of Gaussians the global variables β are the mixture components and mixture(More)
Hamiltonian Monte Carlo (HMC) is a Markov chain Monte Carlo (MCMC) algorithm that avoids the random walk behavior and sensitivity to correlated parameters that plague many MCMC methods by taking a series of steps informed by first-order gradient information. These features allow it to converge to high-dimensional target distributions much more quickly than(More)
We present a hybrid algorithm for Bayesian topic models that combines the efficiency of sparse Gibbs sampling with the scalability of online stochastic inference. We used our algorithm to analyze a corpus of 1.2 million books (33 billion words) with thousands of topics. Our approach reduces the bias of variational inference and generalizes to many Bayesian(More)
Many songs in large music databases are not labeled with semantic tags that could help users sort out the songs they want to listen to from those they do not. If the words that apply to a song can be predicted from audio, then those predictions can be used both to automatically annotate a song with tags, allowing users to get a sense of what qualities(More)
Recent research in machine learning has focused on breaking audio spectrograms into separate sources of sound using latent variable decompositions. These methods require that the number of sources be specified in advance, which is not always possible. To address this problem, we develop Gamma Process Nonnegative Matrix Factorization (GaP-NMF), a Bayesian(More)
Artificial neural networks typically have a fixed, non-linear activation function at each neuron. We have designed a novel form of piecewise linear activation function that is learned independently for each neuron using gradient descent. With this adaptive activation function, we are able to improve upon deep neural network architectures composed of static(More)
Source separation models that make use of nonnegativity in their parameters have been gaining increasing popularity in the last few years, spawning a significant number of publications on the topic. Although these techniques are conceptually similar to other matrix decompositions, they are surprisingly more effective in extracting perceptually meaningful(More)