Oskar Allerbo

Oskar Allerbo, Rebecka Jörnsten

Most machine learning methods require tuning of hyper-parameters. For kernel ridge regression with the Gaussian kernel, the hyper-parameter is the bandwidth. The bandwidth specifies the length scale of the kernel and has to be carefully selected to obtain a model with good generalization. The default methods for bandwidth selection, cross-validation and marginal likelihood maximization, often yield good results, albeit at high computational costs. Inspired by Jacobian regularization, we formulate an approximate expression for how the derivatives of the functions inferred by kernel ridge regression with the Gaussian kernel depend on the kernel bandwidth. We use this expression to propose a closed-form, computationally feather-light, bandwidth selection heuristic, based on controlling the Jacobian. In addition, the Jacobian expression illuminates how the bandwidth selection is a trade-off between the smoothness of the inferred function and the conditioning of the training data kernel matrix. We show on real and synthetic data that compared to cross-validation and marginal likelihood maximization, our method is on pair in terms of model performance, but up to six orders of magnitude faster.

Oskar Allerbo, Thomas B. Schön

We demonstrate how supervised learning can be decomposed into a two-stage procedure, where (1) all model parameters are selected in an unsupervised manner, and (2) the outputs y are added to the model, without changing the parameter values. This is achieved by a new model selection criterion that - in contrast to cross-validation - can be used also without access to y. For linear ridge regression, we bound the asymptotic out-of-sample risk of our method in terms of the optimal asymptotic risk. We also demonstrate that versions of linear and kernel ridge regression, smoothing splines, k-nearest neighbors, random forests, and neural networks, trained without access to y, perform similarly to their standard y-based counterparts. Hence, our results suggest that the difference between supervised and unsupervised learning is less fundamental than it may appear.

Oskar Allerbo, Thomas B. Schön

An accurate assessment of a model's complexity is crucial for topics such as interpretation, generalization, and model selection. However, most existing complexity measures either rely on heuristic assumptions or are computationally prohibitive. In this paper, we present a mathematically rigorous yet easy-to-compute measure of model complexity that is based on the similarities between the model gradients across inputs. It is thus well-defined for any parametric model, but also for kernel-based non-parametric models. We prove that our measure of complexity generalizes model-specific complexity measures such as polynomial degree (for polynomial regression), kernel length scale (for Matérn kernels), number of neighbors (for k-nearest neighbors), number of splits (for decision trees), and number of trees (for random forests). We also use our measure to obtain new insights into the double descent phenomenon for random Fourier features, random forests, neural networks, and gradient boosting.

Oskar Allerbo

Kernel ridge regression, KRR, is a generalization of linear ridge regression that is non-linear in the data, but linear in the model parameters. Here, we introduce an equivalent formulation of the objective function of KRR, which opens up for replacing the ridge penalty with the $\ell_\infty$ and $\ell_1$ penalties. Using the $\ell_\infty$ and $\ell_1$ penalties, we obtain robust and sparse kernel regression, respectively. We study the similarities between explicitly regularized kernel regression and the solutions obtained by early stopping of iterative gradient-based methods, where we connect $\ell_\infty$ regularization to sign gradient descent, $\ell_1$ regularization to forward stagewise regression (also known as coordinate descent), and $\ell_2$ regularization to gradient descent, and, in the last case, theoretically bound for the differences. We exploit the close relations between $\ell_\infty$ regularization and sign gradient descent, and between $\ell_1$ regularization and coordinate descent to propose computationally efficient methods for robust and sparse kernel regression. We finally compare robust kernel regression through sign gradient descent to existing methods for robust kernel regression on five real data sets, demonstrating that our method is one to two orders of magnitude faster, without compromised accuracy.

Oskar Allerbo

We investigate changing the bandwidth of a translational-invariant kernel during training when solving kernel regression with gradient descent. We present a theoretical bound on the out-of-sample generalization error that advocates for decreasing the bandwidth (and thus increasing the model complexity) during training. We further use the bound to show that kernel regression exhibits a double descent behavior when the model complexity is expressed as the minimum allowed bandwidth during training. Decreasing the bandwidth all the way to zero results in benign overfitting, and also circumvents the need for model selection. We demonstrate the double descent behavior on real and synthetic data and also demonstrate that kernel regression with a decreasing bandwidth outperforms that of a constant bandwidth, selected by cross-validation or marginal likelihood maximization. We finally apply our findings to neural networks, demonstrating that by modifying the neural tangent kernel (NTK) during training, making the NTK behave as if its bandwidth were decreasing to zero, we can make the network overfit more benignly, and converge in fewer iterations.

Oskar Allerbo, Rebecka Jörnsten

High-dimensional data sets are often analyzed and explored via the construction of a latent low-dimensional space which enables convenient visualization and efficient predictive modeling or clustering. For complex data structures, linear dimensionality reduction techniques like PCA may not be sufficiently flexible to enable low-dimensional representation. Non-linear dimension reduction techniques, like kernel PCA and autoencoders, suffer from loss of interpretability since each latent variable is dependent of all input dimensions. To address this limitation, we here present path lasso penalized autoencoders. This structured regularization enhances interpretability by penalizing each path through the encoder from an input to a latent variable, thus restricting how many input variables are represented in each latent dimension. Our algorithm uses a group lasso penalty and non-negative matrix factorization to construct a sparse, non-linear latent representation. We compare the path lasso regularized autoencoder to PCA, sparse PCA, autoencoders and sparse autoencoders on real and simulated data sets. We show that the algorithm exhibits much lower reconstruction errors than sparse PCA and parameter-wise lasso regularized autoencoders for low-dimensional representations. Moreover, path lasso representations provide a more accurate reconstruction match, i.e. preserved relative distance between objects in the original and reconstructed spaces.

Oskar Allerbo, Rebecka Jörnsten

Non-parametric, additive models are able to capture complex data dependencies in a flexible, yet interpretable way. However, choosing the format of the additive components often requires non-trivial data exploration. Here, as an alternative, we propose PrAda-net, a one-hidden-layer neural network, trained with proximal gradient descent and adaptive lasso. PrAda-net automatically adjusts the size and architecture of the neural network to reflect the complexity and structure of the data. The compact network obtained by PrAda-net can be translated to additive model components, making it suitable for non-parametric statistical modelling with automatic model selection. We demonstrate PrAda-net on simulated data, where wecompare the test error performance, variable importance and variable subset identification properties of PrAda-net to other lasso-based regularization approaches for neural networks. We also apply PrAda-net to the massive U.K. black smoke data set, to demonstrate how PrAda-net can be used to model complex and heterogeneous data with spatial and temporal components. In contrast to classical, statistical non-parametric approaches, PrAda-net requires no preliminary modeling to select the functional forms of the additive components, yet still results in an interpretable model representation.