Bayesian neural networks and density networks

By David John Cameron MacKay. 1995

Subfields: Probabilistic & Bayesian Methods

How researchers see this paper

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

In Probabilistic & Bayesian Methods

Citations in the corpus

f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization (2016, methodology): “Such probabilistic feedforward neural network models were first considered in [22] and [3], here we call these models generative neural samplers.”
Training generative neural networks via Maximum Mean Discrepancy optimization (2015, background): “The ﬁrst direct instantiation of this idea is due to MacKay [7], although MacKay draws connections even further back to the work of Saund [11] and others on autoencoders, suggesting that generators can be understood as decoders.”
Bayesian Hypernetworks (2017, background): “Bayesian DNNs have been studied since the 1990s (Neal, 1996; MacKay, 1994).”
Hierarchical Gaussian process latent variable models (2007, methodology): “The probabilistic approach to non-linear dimensionality reduction (MacKay, 1995; Bishop et al., 1998) is to formulate a latent variable model, where the la-”
Continuous Sigmoidal Belief Networks Trained using Slice Sampling (1996, methodology): “MacKay (1995) has developed \density networks"that can model both continuous and categorical latent spaces using stochasticity atthe top-most network layer.”
GTM: The Generative Topographic Mapping (1998, methodology): “One way to view the role of the latent space samples {xi} is as a Monte Carlo approximation to the integral over x in (2) (MacKay 1995; Bishop, Svensén, and Williams 1996).”
Semantic Interpolation in Implicit Models (2017, background): “Recently there has been an increased interest in implict models, where a complex generative mechanism is driven by a source of randomness (cf. MacKay (1995)).”
Connectionist multivariate density-estimation and its application to speech synthesis (2016, background): “An extension of our efforts to the Bayesian framework would encounter the usual complications, which are more often computational than conceptual (Neal, 1995; MacKay, 1995a).”
Importance Weighted Hierarchical Variational Inference (2019, background): “…approximation abilities and great empirical success propelled a lot of interest in employing them as powerful sample generators that are trained to output samples from approximate posterior when fed some standard noise as input (Nowozin et al., 2016; Goodfellow et al., 2014; MacKay, 1995).”
Local distance preservation in the GP-LVM through back constraints (2006, background): “One suggested approach is to make use of point based representations of the latent space either through sampling, such as in density networks (MacKay, 1995), or through an explicitly point based representation of the latent space, such as the generative topographic mapping (Bishop et al.”
Bayesian Optimization and Semiparametric Models with Applications to Assistive Technology (2014, methodology): “However, allowing for separate values of each performs Automatic Relevance Determination [MacKay, 1994a] as it allows the model to select the relevant dimensions of the input during learning.”
Fixing Variational Bayes: Deterministic Variational Inference for Bayesian Neural Networks (2018, methodology): “We adopt the paradigm of variational Bayes (VB) for BNNs (Hinton & van Camp, 1993; MacKay, 1995c) which is normally deployed using Monte ∗Work carried out during an internship at Microsoft Research, Cambridge. ar X iv :1 81 0.”
Nonparametric guidance of autoencoder representations using label information (2012, methodology): “A GPLVM with the covariance function of Williams (1998), although it does not impose a density over the data, is similar to a density network (MacKay, 1994) with an infinite number of hidden units in the single hidden layer.”
Pattern Recognition and Machine Learning (2006, unknown): “This is exploited in a general latent variable model called a density network (MacKay, 1995; MacKay and Gibbs, 1999) in which the nonlinear function is governed by a multilayered neural network.”
Learning the Structure of Deep Sparse Graphical Models (2009, background): “This work can be viewed as an infinite multilayer generalization of the density network [MacKay, 1995], and also as part of a more general literature of learning structure in probabilistic networks.”