Pattern Recognition and Machine Learning

By Radford M. Neal. 2006

40,958 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Environmental Engineering, Statistics and Probability, Representation & Self-Supervised Learning

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Environmental Engineering

In Statistics and Probability

In Representation & Self-Supervised Learning

Citations in the corpus

Fully convolutional networks for semantic segmentation (2014, background): “on the standard metrics, their output is dissatisfyingly coarse (see Figure4). The stride of the network prediction limits the scale of detail in the upsampled output. We address this by adding skips [51] that fuse layer outputs, in particular to include shallower layers with ﬁner strides in predict”
Aggregating Local Image Descriptors into Compact Codes (2012, methodology): “Principal component analysis is a standard tool [34] for dimensionality reduction: The eigenvectors associated with the most energetic eigenvalues of the empirical vector covariance matrix are used to define a matrix M mapping a vector x 2 IR(128) to a vector x0 1⁄4Mx in a lower dimensional space.”
Revisiting Natural Gradient for Deep Networks (2013, methodology): “1 Adapting natural gradient descent for neural networks In order to use natural gradient descent for deterministic neural networks we rely on their probabilistic interpretation (see Bishop (2006), chapter 3.”
Inferring Algorithmic Patterns with Stack-Augmented Recurrent Nets (2015, methodology): “This lead to the use of non-linear hidden layers (Rumelhart et al., 1985) and kernels methods (Bishop, 2006).”
Variational Inference with Normalizing Flows (2015, methodology): “Whereas we would have previously resorted to local variational methods (Bishop, 2006), in general we now always compute such expectations using Monte Carlo approximations (including the KL term in the bound, if it is not analytically known).”
NICE: Non-linear Independent Components Estimation (2014, background): “For instance, as highlighted in (Bishop, 2006), a fully parametrized a mixture of gaussians can use one of its mixture component to model one of the datapoints with arbitrary precision, arbitrarily raising the training log-likelihood.”
Semantics derived automatically from language corpora contain human-like biases (2016, background): “In AI and machine learning, bias refers to prior information, a necessary prerequisite for intelligence (Bishop, 2006).”
A note on the evaluation of generative models (2015, background): “…less related to log-likelihood, such as contrastive divergence (Hinton, 2002), score matching (Hyv¨arinen, 2005), lower bounds on the log-likelihood (Bishop, 2006), noise-contrastive estimation (Gutmann & Hyv¨arinen, 2010), probability ﬂow (Sohl-Dickstein et al., 2011), maximum mean discrepancy…”
Meta-learning with differentiable closed-form solvers (2018, positive): “-D2 with the state-of-the-art in few-shot N-class classiﬁcation, it is possible to jointly consider Nbinary classiﬁers, each of which discriminates between a speciﬁc class and all the remaining ones (Bishop, 2006, Chapter 4.1). In our framework, this can be easily implemented by concatenating togeth”
Value Iteration Networks (2016, neutral): “The code for reproducing this results will be made publicly available.”
Virtual Adversarial Training: A Regularization Method for Supervised and Semi-Supervised Learning (2017, background): “From Bayesian standpoint, regularization term can be interpreted as a prior distribution that reflects our educated a priori knowledge or belief regarding the model [7].”
Optimizing Neural Networks with Kronecker-factored Approximate Curvature (2015, methodology): “As this graphical model has a tree structure, there is an equivalent directed graphical model with the same distribution and the same (undirected) graphical structure (e.g. Bishop, 2006), where the directionality of the edges is given by a directed acyclic graph (DAG).”
Black Box Variational Inference (2013, methodology): “On this model, we compared Black Box Variational Inference to Metropolis-Hastings inside Gibbs.”
A Deep and Tractable Density Estimator (2013, neutral): “Bishop, 2006).”
Stochastic variational inference (2012, methodology): “Further, the optimal mean-ﬁeld distribution, without regard to its particular functional form, has factors in these families (Bishop, 2006).”
Self-labelling via simultaneous clustering and representation learning (2019, background): “However, combining representation learning, which is a discriminative task, with clustering, which is usually based on a generative objective (Bishop, 2006), is not at all trivial.”