Abstract:We introduce a new regularization method for Artificial Neural Networks (ANNs) based on Kernel Flows (KFs). KFs were introduced as a method for kernel selection in regression/kriging based on the minimization of the loss of accuracy incurred by halving the number of interpolation points in random batches of the dataset. Writing $f_\theta(x) = \big(f^{(n)}_{\theta_n}\circ f^{(n-1)}_{\theta_{n-1}} \circ \dots \circ f^{(1)}_{\theta_1}\big)(x)$ for the functional representation of compositional structure of the ANN, the inner layers outputs $h^{(i)}(x) = \big(f^{(i)}_{\theta_i}\circ f^{(i-1)}_{\theta_{i-1}} \circ \dots \circ f^{(1)}_{\theta_1}\big)(x)$ define a hierarchy of feature maps and kernels $k^{(i)}(x,x')=\exp(- \gamma_i \|h^{(i)}(x)-h^{(i)}(x')\|_2^2)$. When combined with a batch of the dataset these kernels produce KF losses $e_2^{(i)}$ (the $L^2$ regression error incurred by using a random half of the batch to predict the other half) depending on parameters of inner layers $\theta_1,\ldots,\theta_i$ (and $\gamma_i$). The proposed method simply consists in aggregating a subset of these KF losses with a classical output loss. We test the proposed method on CNNs and WRNs without alteration of structure nor output classifier and report reduced test errors, decreased generalization gaps, and increased robustness to distribution shift without significant increase in computational complexity. We suspect that these results might be explained by the fact that while conventional training only employs a linear functional (a generalized moment) of the empirical distribution defined by the dataset and can be prone to trapping in the Neural Tangent Kernel regime (under over-parameterizations), the proposed loss function (defined as a nonlinear functional of the empirical distribution) effectively trains the underlying kernel defined by the CNN beyond regressing the data with that kernel.

Robustly Train Normalizing Flows Via KL Divergence Regularization

Stable Training of Normalizing Flows for High-dimensional Variational Inference

Normalizing flow sampling with Langevin dynamics in the latent space

Deep regularization and direct training of the inner layers of Neural Networks with Kernel Flows

Learning Likelihoods with Conditional Normalizing Flows

Normalizing Flow-Based Probability Distribution Representation Detector for Hyperspectral Anomaly Detection.

Convergence of Continuous Normalizing Flows for Learning Probability Distributions

Normalizing flow regularization for photoacoustic tomography

Understanding Likelihood of Normalizing Flow and Image Complexity through the Lens of Out-of-Distribution Detection

Flows for Flows: Training Normalizing Flows Between Arbitrary Distributions with Maximum Likelihood Estimation

Accelerating Continuous Normalizing Flow with Trajectory Polynomial Regularization

Kullback-Leibler Divergence-Based Out-of-Distribution Detection with Flow-Based Generative Models

Normalizing Flow with Variational Latent Representation

AdvNF: Reducing mode collapse in conditional normalising flows using adversarial learning

Adaptive Epsilon Adversarial Training for Robust Gravitational Wave Parameter Estimation Using Normalizing Flows

Preconditioned training of normalizing flows for variational inference in inverse problems

Flexible Tails for Normalizing Flows

Normalizing Flows for Conformal Regression

ManiFlow: Implicitly Representing Manifolds with Normalizing Flows