David Atienza

Christodoulos Kechris, Jonathan Dan, Jose Miranda et al.

Non-linear activation functions are crucial in Convolutional Neural Networks. However, until now they have not been well described in the frequency domain. In this work, we study the spectral behavior of ReLU, a popular activation function. We use the ReLU's Taylor expansion to derive its frequency domain behavior. We demonstrate that ReLU introduces higher frequency oscillations in the signal and a constant DC component. Furthermore, we investigate the importance of this DC component, where we demonstrate that it helps the model extract meaningful features related to the input frequency content. We accompany our theoretical derivations with experiments and real-world examples. First, we numerically validate our frequency response model. Then we observe ReLU's spectral behavior on two example models and a real-world one. Finally, we experimentally investigate the role of the DC component introduced by ReLU in the CNN's representations. Our results indicate that the DC helps to converge to a weight configuration that is close to the initial random weights.

Christodoulos Kechris, Jonathan Dan, David Atienza

Traditional saliency map methods, popularized in computer vision, highlight individual points (pixels) of the input that contribute the most to the model's output. However, in time-series they offer limited insights as semantically meaningful features are often found in other domains. We introduce Cross-domain Integrated Gradients, a generalization of Integrated Gradients. Our method enables feature attributions on any domain that can be formulated as an invertible, differentiable transformation of the time domain. Crucially, our derivation extends the original Integrated Gradients into the complex domain, enabling frequency-based attributions. We provide the necessary theoretical guarantees, namely, path independence and completeness. Our approach reveals interpretable, problem-specific attributions that time-domain methods cannot capture, on three real-world tasks: wearable sensor heart rate extraction, electroencephalography-based seizure detection, and zero-shot time-series forecasting. We release an open-source Tensorflow/PyTorch library to enable plug-and-play cross-domain explainability for time-series models. These results demonstrate the ability of cross-domain integrated gradients to provide semantically meaningful insights in time-series models that are impossible with traditional time-domain saliency.

Christodoulos Kechris, Jonathan Dan, Jose Miranda et al.

Deep learning time-series processing often relies on convolutional neural networks with overlapping windows. This overlap allows the network to produce an output faster than the window length. However, it introduces additional computations. This work explores the potential to optimize computational efficiency during inference by exploiting convolution's shift-invariance properties to skip the calculation of layer activations between successive overlapping windows. Although convolutions are shift-invariant, zero-padding and pooling operations, widely used in such networks, are not efficient and complicate efficient streaming inference. We introduce StreamiNNC, a strategy to deploy Convolutional Neural Networks for online streaming inference. We explore the adverse effects of zero padding and pooling on the accuracy of streaming inference, deriving theoretical error upper bounds for pooling during streaming. We address these limitations by proposing signal padding and pooling alignment and provide guidelines for designing and deploying models for StreamiNNC. We validate our method in simulated data and on three real-world biomedical signal processing applications. StreamiNNC achieves a low deviation between streaming output and normal inference for all three networks (2.03 - 3.55% NRMSE). This work demonstrates that it is possible to linearly speed up the inference of streaming CNNs processing overlapping windows, negating the additional computation typically incurred by overlapping windows.

David Atienza, Concha Bielza, Pedro Larrañaga

We introduce semiparametric Bayesian networks that combine parametric and nonparametric conditional probability distributions. Their aim is to incorporate the advantages of both components: the bounded complexity of parametric models and the flexibility of nonparametric ones. We demonstrate that semiparametric Bayesian networks generalize two well-known types of Bayesian networks: Gaussian Bayesian networks and kernel density estimation Bayesian networks. For this purpose, we consider two different conditional probability distributions required in a semiparametric Bayesian network. In addition, we present modifications of two well-known algorithms (greedy hill-climbing and PC) to learn the structure of a semiparametric Bayesian network from data. To realize this, we employ a score function based on cross-validation. In addition, using a validation dataset, we apply an early-stopping criterion to avoid overfitting. To evaluate the applicability of the proposed algorithm, we conduct an exhaustive experiment on synthetic data sampled by mixing linear and nonlinear functions, multivariate normal data sampled from Gaussian Bayesian networks, real data from the UCI repository, and bearings degradation data. As a result of this experiment, we conclude that the proposed algorithm accurately learns the combination of parametric and nonparametric components, while achieving a performance comparable with those provided by state-of-the-art methods.