Search

The 5G Radio Access Network (RAN) virtualization aims to improve network quality and lower the operator's costs. One of its main features is the functional split, i.e., dividing the instantiation of RAN baseband functions into different units over metro-network nodes. However, its optimal placement is non-trivial: it depends on the application requirements and on the expected traffic volume, whose daily variation highly impacts the total power consumption. Current optimization solutions fail to provide a placement solution capable of handling traffic fluctuations. In fact, the standard machine learning algorithms used in the literature for planning the network resources in advance result in an allocation that is inadequate to carry the actual traffic at all the time-slots. Hence, we must reserve an artificial buffer capacity in the nodes to ensure feasibility. Instead, our proposed method exploits a fine-grained two-step multi-task algorithm that predicts the mean and quantile traffic, making the artificial capacity no longer necessary. The subsequent placement uses mixed-integer linear programming and a heuristic. The former considers the expected traffic in the objective function (to estimate costs) and the quantile in the constraints (to enforce capacity limits). The heuristic combines the mean and quantile results to minimize the power and comply with the requirements. While using sufficiently large artificial buffers guarantees robustness with a mild power increase compared to the oracle, the fine-grained multi-task model improves the results, reducing the power consumption compared to the mean and meets all constraints. The heuristic enables significant computational time reduction., Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public., Cyber Security

One method to solve expensive black-box optimization problems is to use a surrogate model that approximates the objective based on previous observed evaluations. The surrogate, which is cheaper to evaluate, is optimized instead to find an approximate solution to the original problem. In the case of discrete problems, recent research has revolved around discrete surrogate models that are specifically constructed to deal with these problems. A main motivation is that literature considers continuous methods, such as Bayesian optimization with Gaussian processes as the surrogate, to be sub-optimal (especially in higher dimensions) because they ignore the discrete structure by, e.g., rounding off real-valued solutions to integers. However, we claim that this is not true. In fact, we present empirical evidence showing that the use of continuous surrogate models displays competitive performance on a set of high-dimensional discrete benchmark problems, including a real-life application, against state-of-the-art discrete surrogate-based methods. Our experiments with different kinds of discrete decision variables and time constraints also give more insight into which algorithms work well on which type of problem., Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public., Pattern Recognition and Bioinformatics, Algorithmics, Cyber Security

A challenging problem in both engineering and computer science is that of minimising a function for which we have no mathematical formulation available, that is expensive to evaluate, and that contains continuous and integer variables, for example in automatic algorithm configuration. Surrogate-based algorithms are very suitable for this type of problem, but most existing techniques are designed with only continuous or only discrete variables in mind. Mixed-Variable ReLU-based Surrogate Modelling (MVRSM) is a surrogate-based algorithm that uses a linear combination of rectified linear units, defined in such a way that (local) optima satisfy the integer constraints. Unlike other methods, it also has a constant run-time per iteration. This method outperforms the state of the art on several synthetic benchmarks with up to 238 continuous and integer variables, and achieves competitive performance on two real-life benchmarks: XG-Boost hyperparameter tuning and Electrostatic Precipitator optimisation., Cyber Security, Algorithmics

The industrial challenge of the GECCO 2021 conference is an expensive optimisation problem, where the parameters of a hospital simulation model need to be tuned to optimality. We show how a surrogate-based optimisation framework, with a random ReLU expansion as the surrogate model, outperforms other methods such as Bayesian optimisation, Hyperopt, and random search on this problem.

When a black-box optimization objective can only be evaluated with costly or noisy measurements, most standard optimization algorithms are unsuited to find the optimal solution. Specialized algorithms that deal with exactly this situation make use of surrogate models. These models are usually continuous and smooth, which is beneficial for continuous optimization problems, but not necessarily for combinatorial problems. However, by choosing the basis functions of the surrogate model in a certain way, we show that it can be guaranteed that the optimal solution of the surrogate model is integer. This approach outperforms random search, simulated annealing and a Bayesian optimization algorithm on the problem of finding robust routes for a noise-perturbed traveling salesman benchmark problem, with similar performance as another Bayesian optimization algorithm, and outperforms all compared algorithms on a convex binary optimization problem with a large number of variables., Algorithmics, Cyber Security

Training classifiers that are robust against adversarially modified examples is becoming increasingly important in practice. In the field of malware detection, adversaries modify malicious binary files to seem benign while preserving their malicious behavior. We report on the results of a recently held robust malware detection challenge. There were two tracks in which teams could participate: the attack track asked for adversarially modified malware samples and the defend track asked for trained neural network classifiers that are robust to such modifications. The teams were unaware of the attacks/defenses they had to detect/evade. Although only 9 teams participated, this unique setting allowed us to make several interesting observations. We also present the challenge winner: GRAMS, a family of novel techniques to train adversarially robust networks that preserve the intended (malicious) functionality and yield high-quality adversarial samples. These samples are used to iteratively train a robust classifier. We show that our techniques, based on discrete optimization techniques, beat purely gradient-based methods. GRAMS obtained first place in both the attack and defend tracks of the competition., Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public., Cyber Security, Algorithmics

Beamforming is a signal processing technique used in highly directional antennas. An array of antenna elements transmits the same signal, but with a different time delay for each element. By providing the right time delays for each antenna element, the whole array transmits a high-powered signal in one desired direction. This technique can be used for example to provide satellite television and Internet connections on board of aircrafts. Recently, developments in the field of integrated microwave photonics have paved the way for broadband, low-loss, and low-weight beamformer systems. These photonic beamformers convert the signals to be transmitted to the optical domain, provide the correct time delays with tunable optical delay lines, and then convert the signal back to the radio frequency domain. The main challenge here lies in tuning the actuators of the tunable optical delay lines in such a way that they provide the desired time delays. Challenges like actuator crosstalk, parameter sensitivity, noise and model errors cause complications when traditional tuning algorithms are used, such as nonlinear optimization routines. All results obtained with these photonic beamformers in the literature so far have been achieved by tuning the whole system by hand, or by applying nonlinear optimization techniques to a simplified simulation of the system rather than the actual system. In order to find a practical way of tuning a photonic beamformer in real time, this thesis takes a data-driven approach. Instead of relying on perfectly accurate physical models, a surrogate function is used that approximates the relation between the system actuators and a cost function, namely the difference between the measured and desired time delay of each antenna element. By performing nonlinear optimization techniques on this surrogate cost function and by continuously updating the approximation as new measurements are obtained, the time delays of each antenna element should converge tow, Numerics for Control & Identification

This paper discusses various techniques to let an agent learn how to predict the effects of its own actions on its sensor data autonomously, and their usefulness to apply them to visual sensors. An Extreme Learning Machine is used for visuomotor prediction, while various autonomous control techniques that can aid the prediction process by balancing exploration and exploitation are discussed and tested in a simple system: a camera moving over a 2D greyscale image.

Most robots receive information about their own dynamics or about their environment before starting to work on their tasks. This makes them less fit for operating in unknown environments without human supervision. In this thesis, the goal is to let a robot learn the relation between its actions and perceptions autonomously. This problem is divided in a prediction part, and a control part. The goal of the prediction process is to correctly predict the sensor data of the future, given the past sensor and actuator data. The goal of the controller is to provide exploratory behaviour, so that different situations can be learned by the predictor. Of the discussed techniques, one predictor and one controller are chosen to be used in an experiment: a mobile robot equipped with a camera that has to learn what happens when rotating inside a box with pictures. The chosen predictor is a new type of local linear model, the Shuffled Linear Model (SLM). In this model, the localities are chosen randomly and linear regression is used to determine the linear model parameters. It is proved that this model is a universal approximator, and an upper bound on the number of models is given. The chosen controller makes use of intrinsic rewards to aid the predictor by providing data that is both new and easy to predict., Applied mathematics, Electrical Engineering, Mathematics and Computer Science

The memristor is an electronic component which was invented in theory by Leon Chua in 1971, based on logics and symmetry. In 2008, HP Labs has realized an actual memristor on the nanoscale level for the first time in history. Since then, a lot of researchers have focused their attention on the memristor. It is one of the fundamental passive circuit elements, like the resistor, the capacitor and the inductor. What differentiates the memristor from these other fundamental passive circuit elements is its most useful property: it has a memory. Aside from this, there are a lot of similarities between the memristor and the synapses in our brain, which makes it a very useful research subject in the area of artificial intelligence, especially neural networks. These networks are known for their ability to learn, and in this thesis the mathematics behind this learning process will be explained., Optimalisering en systeemtheorie, Electrical Engineering, Mathematics and Computer Science