Learning for Larger Datasets with the Gaussian Process Latent Variable Model

By Neil D. Lawrence. 2007

129 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Training energy-based models for time-series imputation (2013, background): “There has been some work on improving the computational efficiency of these models by introducing sparsity (Lawrence, 2007) but parametric graphical models tend to remain more practical for larger data sets.”
Two Distributed-State Models For Generating High-Dimensional Time Series (2011, methodology): “Such approaches have produced impressive results given user-suppliedmotion content but we seek more powerful methods that can synthesize both style and content.”
Computationally Efficient Convolved Multiple Output Gaussian Processes (2011, methodology): “We follow the notation of Brookes (2005) obtaining similar results to Lawrence(2007).”
Factored conditional restricted Boltzmann Machines for modeling motion style (2009, background): “The application does not naturally present a quantitative comparison, but in the past, other time series models have been compared by their performance on the prediction of either full or partial held-out frames (Taylor et al., 2007; Lawrence, 2007).”
Variational Gaussian Process Dynamical Systems (2011, methodology): “The data set was constructed and preprocessed as described in [15].”
Geodesic Finite Mixture Models (2014, background): “Again, despite some approximations to improve computation efficiency [19, 24], the algorithmic complexity of GP makes them unfeasible for large datasets, as those we consider here.”
Bayesian Gaussian Process Latent Variable Model (2010, methodology): “We follow the approach of Lawrence (2007): for each vector of latent function values f d we introduce a separate set of M inducing variables u d ∈ Ê M evaluated at a set of inducing input locations given by Z ∈ Ê M ×Q .”
State of the Art on Monocular 3D Face Reconstruction, Tracking, and Applications (2018, methodology): “Finally, also non-linear face models can be applied [Law07] for face capture.”