Refine
Document Type
- Conference Proceeding (12) (remove)
Conference Type
- Konferenzartikel (12)
Language
- English (12) (remove)
Is part of the Bibliography
- yes (12)
Keywords
- Machine Learning (12) (remove)
Institute
- Fakultät Elektrotechnik, Medizintechnik und Informatik (EMI) (ab 04/2019) (9)
- IMLA - Institute for Machine Learning and Analytics (7)
- Fakultät Elektrotechnik und Informationstechnik (E+I) (bis 03/2019) (1)
- Fakultät Maschinenbau und Verfahrenstechnik (M+V) (1)
- Fakultät Medien (M) (ab 22.04.2021) (1)
- INES - Institut für nachhaltige Energiesysteme (1)
Open Access
- Open Access (9)
- Bronze (4)
- Closed (2)
- Closed Access (1)
- Diamond (1)
Polyarticulated active prostheses constitute a promising solution for upper limb amputees. The bottleneck for their adoption though, is the lack of intuitive control. In this context, machine learning algorithms based on pattern recognition from electromyographic (EMG) signals represent a great opportunity for naturally operating prosthetic devices, but their performance is strongly affected by the selection of input features. In this study, we investigated different combinations of 13 EMG-derived features obtained from EMG signals of healthy individuals performing upper limb movements and tested their performance for movement classification using an Artificial Neural Network. We found that input data (i.e., the set of input features) can be reduced by more than 50% without any loss in accuracy, while diminishing the computing time required to train the classifier. Our results indicate that input features must be properly selected in order to optimize prosthetic control.
Soiling is an important issue in the renewable energy sector since it can result in significant yield losses, especially in regions with higher pollution or dust levels. To mitigate the impact of soiling on photovoltaic (PV) plants, it is essential to regularly monitor and clean the panels, as well as develop accurate soiling predictions that can affect cleaning strategies and enhance the overall performance of PV power plants. This research focuses on the problem of soiling loss in photovoltaic power plants and the potential to improve the accuracy of soiling predictions. The study examines how soiling can affect the efficiency and productivity of the modules and how to measure and predict soiling using machine learning (ML) algorithms. The research includes analyzing real data from large-scale ground-mounted PV sites and comparing different soiling measurement methods. It was observed that there were some deviations in the real soiling loss values compared to the expected values for some projects in southern Spain, thus, the main goal of this work is to develop machine learning models that could predict the soiling more accurately. The developed models have a low mean square error (MSE), indicating the accuracy and suitability of the models to predict the soiling rates. The study also investigates the impact of different cleaning strategies on the performance of PV power plants and provides a powerful application to predict both the soiling and the number of cleaning cycles.
Recent work has investigated the distributions of learned convolution filters through a large-scale study containing hundreds of heterogeneous image models. Surprisingly, on average, the distributions only show minor drifts in comparisons of various studied dimensions including the learned task, image domain, or dataset. However, among the studied image domains, medical imaging models appeared to show significant outliers through "spikey" distributions, and, therefore, learn clusters of highly specific filters different from other domains. Following this observation, we study the collected medical imaging models in more detail. We show that instead of fundamental differences, the outliers are due to specific processing in some architectures. Quite the contrary, for standardized architectures, we find that models trained on medical data do not significantly differ in their filter distributions from similar architectures trained on data from other domains. Our conclusions reinforce previous hypotheses stating that pre-training of imaging models can be done with any kind of diverse image data.
Harnessing the overall benefits of the latest advancements in artificial intelligence (AI) requires the extensive collaboration of academia and industry. These collaborations promote innovation and growth while enforcing the practical usefulness of newer technologies in real life. The purpose of this article is to outline the challenges faced during cross-collaboration between academia and industry. These challenges are also inspected with the help of an ongoing project titled “Quality Assurance of Machine Learning Applications” (Q-AMeLiA), in which three universities cooperate with five industry partners to make the product risk of AI-based products visible. Further, we discuss the hurdles and the key challenges in machine learning (ML) technology transformation from academia to industry based on robustness, simplicity, and safety. These challenges are an outcome of the lack of common standards, metrics, and missing regulatory considerations when state-of-the-art (SOTA) technology is developed in academia. The use of biased datasets involves ethical concerns that might lead to unfair outcomes when the ML model is deployed in production. The advancement of AI in small and medium sized enterprises (SMEs) requires more in terms of common tandardization of concepts rather than algorithm breakthroughs. In this paper, in addition to the general challenges, we also discuss domain specific barriers for five different domains i.e., object detection, hardware benchmarking, continual learning, action recognition, and industrial process automation, and highlight the steps necessary for successfully managing the cross-sectoral collaborations between academia and industry.
Recently, RobustBench (Croce et al. 2020) has become a widely recognized benchmark for the adversarial robustness of image
classification networks. In it’s most commonly reported sub-task, RobustBench evaluates and ranks the adversarial robustness of trained neural networks on CIFAR10 under AutoAttack (Croce and Hein 2020b) with l∞ perturbations limited to ϵ = 8/255. With leading scores of the currently best performing models of around 60% of the baseline, it is fair to characterize this benchmark to be quite challenging. Despite it’s general acceptance in recent literature, we aim to foster discussion about the suitability of RobustBench as a key indicator for robustness which could be generalized to practical applications. Our line of argumentation against this is two-fold and supported by excessive experiments presented in this paper: We argue that I) the alternation of data by AutoAttack with l∞, ϵ = 8/255 is unrealistically strong, resulting in close to perfect detection rates of adversarial samples even by simple detection algorithms and human observers.
We also show that other attack methods are much harder to detect while achieving similar success rates. II) That results on low resolution data sets like CIFAR10 do not generalize well to higher resolution images as gradient based attacks appear to become even more detectable with increasing resolutions.
Many commonly well-performing convolutional neural network models have shown to be susceptible to input data perturbations, indicating a low model robustness. Adversarial attacks are thereby specifically optimized to reveal model weaknesses, by generating small, barely perceivable image perturbations that flip the model prediction. Robustness against attacks can be gained for example by using adversarial examples during training, which effectively reduces the measurable model attackability. In contrast, research on analyzing the source of a model’s vulnerability is scarce. In this paper, we analyze adversarially trained, robust models in the context of a specifically suspicious network operation, the downsampling layer, and provide evidence that robust models have learned to downsample more accurately and suffer significantly less from aliasing than baseline models.
Estimating the Robustness of Classification Models by the Structure of the Learned Feature-Space
(2022)
Over the last decade, the development of deep image classification networks has mostly been driven by the search for the best performance in terms of classification accuracy on standardized benchmarks like ImageNet. More recently, this focus has been expanded by the notion of model robustness, \ie the generalization abilities of models towards previously unseen changes in the data distribution. While new benchmarks, like ImageNet-C, have been introduced to measure robustness properties, we argue that fixed testsets are only able to capture a small portion of possible data variations and are thus limited and prone to generate new overfitted solutions. To overcome these drawbacks, we suggest to estimate the robustness of a model directly from the structure of its learned feature-space. We introduce robustness indicators which are obtained via unsupervised clustering of latent representations from a trained classifier and show very high correlations to the model performance on corrupted test data.
The present work ties in with the problem of bicycle road assessment that is currently done using expensive special measuring vehicles. Our alternative approach for road condition assessment is to mount a sensor device on a bicycle which sends accelerometer and gyroscope data via WiFi to a classification server. There, a prediction model determines road type and condition based on the sensor data. For the classification task, we compare different machine learning methods with each other, whereby validation accuracies of 99% can be achieved with deep residual networks such as InceptionTime. The main contribution of this work with respect to comparable work is that we achieve excellent accuracies on a realistic dataset classifying road conditions into nine distinct classes that are highly relevant for practice.
A fundamental and still largely unsolved question in the context of Generative Adversarial Networks is whether they are truly able to capture the real data distribution and, consequently, to sample from it. In particular, the multidimensional nature of image distributions leads to a complex evaluation of the diversity of GAN distributions. Existing approaches provide only a partial understanding of this issue, leaving the question unanswered. In this work, we introduce a loop-training scheme for the systematic investigation of observable shifts between the distributions of real training data and GAN generated data. Additionally, we introduce several bounded measures for distribution shifts, which are both easy to compute and to interpret. Overall, the combination of these methods allows an explorative investigation of innate limitations of current GAN algorithms. Our experiments on different data-sets and multiple state-of-the-art GAN architectures show large shifts between input and output distributions, showing that existing theoretical guarantees towards the convergence of output distributions appear not to be holding in practice.
In the field of network security, the detection of possible intrusions is an important task to prevent and analyse attacks. Machine learning has been adopted as a particular supporting technique over the last years. However, the majority of related published work uses post mortem log files and fails to address the required real-time capabilities of network data feature extraction and machine learning based analysis [1-5]. We introduce the network feature extractor library FEX, which is designed to allow real-time feature extraction of network data. This library incorporates 83 statistical features based on reassembled data flows. The introduced Cython implementation allows processing individual packets within 4.58 microseconds. Based on the features extracted by FEX, existing intrusion detection machine learning models were examined with respect to their real-time capabilities. An identified Decision-Tree Classifier model was thus further optimised by transpiling it into C Code. This reduced the prediction time of a single sample to 3.96 microseconds on average. Based on the feature extractor and the improved machine learning model an IDS system was implemented which supports a data throughput between 63.7 Mbit/s and 2.5 Gbit/s making it a suitable candidate for a real-time, machine-learning based IDS.