In this theory paper, we investigate training deep neural networks (DNNs) for classification via minimizing the information bottleneck (IB) functional. We show that resulting optimization problem suffers from two severe issues: First, deterministic DNNs, either IB functional is infinite almost all values of network parameters, making ill-posed, or it piecewise constant, hence not admitting gradient-based methods. Second, invariance under bijections prevents capturing properties learned...

Finite precision approximations of discrete probability distributions are considered, applicable for distribution synthesis, e.g., probabilistic shaping. Two algorithms presented that find the optimal $M$-type approximation $Q$ a $P$ in terms variational distance $| Q-P|_1$ and informational divergence $\mathbb{D}(Q| P)$. Bounds on errors derived shown to be asymptotically tight. Several examples illustrate can quite different from approximation.

Digital product passports (DPPs) are an emerging technology and considered as enablers of sustainable circular value chains they support management (SPM) by gathering containing life cycle data. However, some data sensitive stakeholders, resulting in a reluctance to share such This contribution provides concept illustrating how science machine learning approaches enable electric vehicle battery (EVB) chain stakeholders carry out confidentiality-preserving exchange via DPP. This, turn, can...

The technical world of today fundamentally relies on structural analysis in the form design and mechanic simulations. A traditional robust simulation method is physics-based finite element (FEM) simulation. FEM simulations mechanics are known to be very accurate; however, higher desired resolution, more computational effort required. Surrogate modeling provides a approach address this drawback. Nonetheless, finding right surrogate model its hyperparameters for specific use case not...

Article Many-Objective Simulation Optimization for Camp Location Problems in Humanitarian Logistics Yani Xue 1,*, Miqing Li 2, Hamid Arabnejad 1, Diana Suleimenova Alireza Jahani Bernhard C. Geiger 3, Freek Boesjes 4, Anastasia Anagnostou Simon J.E. Taylor Xiaohui Liu and Derek Groen 1,* 1 Department of Computer Science, Brunel University London, Uxbridge, United Kingdom 2 School Birmingham, 3 Know-Center GmbH, Graz, Austria 4 Faculty Geosciences, Utrecht University, Utrecht, The Netherlands...

In this paper, we consider a lossy single-user caching problem with correlated sources. We first describe the fundamental interplay between source correlations, capacity of user's cache, reconstruction distortion requirements, and final delivery-phase (compression) rate. then illustrate using multivariate Gaussian example binary symmetric example. To fully explore effect formulate f-separable functions recently introduce by Shkel Verdú. The class includes separable as special case, our...

In this paper, the authors investigated changes in mass concentrations of particulate matter (PM) during Coronavirus Disease 2019 (COVID-19) lockdown. Daily samples PM

Physics-informed neural networks (PINNs) have emerged as a promising deep learning method, capable of solving forward and inverse problems governed by differential equations. Despite their recent advance, it is widely acknowledged that PINNs are difficult to train often require careful tuning loss weights when data physics functions combined scalarization multi-objective (MO) problem. In this paper, we aim understand how parameters the physical system, such characteristic length time scales,...

Understanding the functional architecture of complex systems is crucial to illuminate their inner workings and enable effective methods for prediction control. Recent advances have introduced tools characterise emergent macroscopic levels; however, while these approaches are successful in identifying when emergence takes place, they limited extent can determine how it does. Here we address this limitation by developing a computational approach emergence, which characterises processes terms...

The phenomenon of knock is an abnormal combustion occurring in spark-ignition (SI) engines and forms a barrier that prevents increase thermal efficiency while simultaneously reducing CO2 emissions. Since knocking highly stochastic, cyclic analysis in-cylinder pressure necessary. In this study we propose approach for efficient robust detection identification three different internal engines. proposed methodology includes signal processing technique, called continuous wavelet transformation...

This paper introduces a method for the detection of knock occurrences in an internal combustion engine (ICE) using 1D convolutional neural network trained on in-cylinder pressure data. The model architecture was based considerations regarding expected frequency characteristics knocking combustion. To aid feature extraction, all cycles were reduced to 60{\deg} CA long windows, with no further processing applied traces. networks exclusively traces from multiple conditions and labels provided...

Abstract Most engineering domains abound with models derived from first principles that have beenproven to be effective for decades. These are not only a valuable source of knowledge, but they also form the basis simulations. The recent trend digitization has complemented these data in all forms and variants, such as process monitoring time series, measured material characteristics, stored production parameters. Theory-inspired machine learning combines available data, reaping benefits...

Superficially, read and spontaneous speech—the two main kinds of training data for automatic speech recognition—appear as complementary, but are equal: pairs texts acoustic signals. Yet, is typically harder recognition. This usually explained by different variation noise, there a more fundamental deviation at play: speech, the audio signal produced recitation given text, whereas in text transcribed from signal. In this review, we embrace difference presenting first introduction causal...

A comparison between two widespread global navigation satellite system (GNSS) acquisition strategies is presented. The first strategy bases its decision on comparing the energy within a cell to threshold, while second one uses ratio largest energies. It shown that method outperforms in terms of receiver operating characteristics (ROCs) many practically relevant cases. Moreover, despite purported simplicity detection method, it further complexity comparable or even higher than threshold with...

In this work we analyze principle component analysis (PCA) as a deterministic input-output system. We show that the relative information loss induced by reducing dimensionality of data after performing PCA is same in reduction without PCA. Furthermore, case where uses sample covariance matrix to compute rotation. If rotation not available at output, an infinite amount lost. The shown decrease with increasing size.

In 1959, R\'enyi proposed the information dimension and $d$-dimensional entropy to measure content of general random variables. This paper proposes a generalization stochastic processes by defining rate as uniformly-quantized process divided minus logarithm quantizer step size $1/m$ in limit $m\to\infty$. It is demonstrated that coincides with rate-distortion dimension, defined twice function $R(D)$ $-\log(D)$ $D\downarrow 0$. further shown that, among all multivariate stationary given...

Abstract Complex systems, abstractly represented as networks, are ubiquitous in everyday life. Analyzing and understanding these systems requires, among others, tools for community detection. As no single best detection algorithm can exist, robustness across a wide variety of problem settings is desirable. In this work, we present Synwalk, random walk-based method. Synwalk builds upon solid theoretical basis detects communities by synthesizing the walk induced given network from class...

The avoidance of scrap and the adherence to tolerances is an important goal in manufacturing. This requires a good engineering understanding underlying process. To achieve this, real physical experiments can be conducted. However, they are expensive time resources, slow down production. A promising way overcome these drawbacks process exploration through simulation, where finite element method (FEM) well-established robust simulation method. While FEM provide high-resolution results, it...

This paper empirically studies commonly observed training difficulties of Physics-Informed Neural Networks (PINNs) on dynamical systems. Our results indicate that fixed points which are inherent to these systems play a key role in the optimization PINNs embedded physics loss function. We observe landscape exhibits local optima shaped by presence points. find contribute complexity can explain common and resulting nonphysical predictions. Under certain settings, e.g., initial conditions close...

An optimal control of the combustion process an engine ensures lower emissions and fuel consumption plus high efficiencies. Combustion parameters such as peak firing pressure (PFP) crank angle (CA) corresponding to 50% mass fraction burned (MFB50) are essential for a closed-loop strategy. These based on measured in-cylinder that is typically gained by intrusive sensors (PSs). costly their durability uncertain. To overcome these issues, potential using virtual sensor vibration signals...

The great potential of free space optics (FSO) communication motivates to use it for future requirement high bandwidth links. However, the widespread this technology has been hampered by reduced availability because weather effects. To overcome these shortcomings a back-up link can be used so that combined hybrid network may achieve carrier class availability. A hybrid, FSO/wireless LAN (WLAN) self-synchronising architecture is presented, which provides transparent connectivity without...