ACCURACY VERSUS INTERPRETABILITY IN FLEXIBLE MODELING : IMPLEMENTING A TRADEOFF USING GAUSSIAN PROCESS MODELS

How researchers see this paper

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

Learning customized and optimized lists of rules with mathematical programming (2018, background): “The goal of designing interpretable classifiers has existed since the beginning of artificial intelligence, (see reviews of [21,31,52,59,68]).”
Structure Discovery in Nonparametric Regression through Compositional Kernel Search (2013, background): “Plate (1999) constructs a gp with a composite kernel, summing an SE kernel along each dimension, with an SE-ARD kernel (i.e. a product of SE over all dimensions).”
Additive Approximations in High Dimensional Nonparametric Regression via the SALSA (2016, methodology): “While at first sight the differences with KRR might seem superficial, we will see that the stronger additive assumption will help us reduce the excess risk for high dimensional regression.”
Neural network explanation using inversion (2007, background): “Other algorithms which addressed a tradeoff between interpretability and accuracy include that of Plate (1999).”
Structural Modelling with Sparse Kernels (2002, background): “This is especially true in medical domains, where black-box models, such as traditional neural networks, bagged descision trees as well as kernel methods, are viewed with great suspicion (Wyatt, 1995; Plate, 1999).”
Additive Gaussian Processes (2011, methodology): “Plate [6] constructs a form of additive GP, but using only the ﬁrst-order and D th order terms.”