Refine
Year of publication
- 2020 (10)
Document Type
- Article (9)
- Conference Proceeding (1)
Language
- English (10)
Keywords
Institute
- Grundlagen (10) (remove)
Over the past few years, deep neural networks have shown excellent results in multiple tasks, however, there is still an increasing need to address the problem of interpretability to improve model transparency, performance, and safety. Achieving eXplainable Artificial Intelligence (XAI) by combining neural networks with continuous logic and multi-criteria decision-making tools is one of the most promising ways to approach this problem: by this combination, the black-box nature of neural models can be reduced. The continuous logic-based neural model uses so-called Squashing activation functions, a parametric family of functions that satisfy natural invariance requirements and contain rectified linear units as a particular case. This work demonstrates the first benchmark tests that measure the performance of Squashing functions in neural networks. Three experiments were carried out to examine their usability and a comparison with the most popular activation functions was made for five different network types. The performance was determined by measuring the accuracy, loss, and time per epoch. These experiments and the conducted benchmarks have proven that the use of Squashing functions is possible and similar in performance to conventional activation functions. Moreover, a further experiment was conducted by implementing nilpotent logical gates to demonstrate how simple classification tasks can be solved successfully and with high performance. The results indicate that due to the embedded nilpotent logical operators and the differentiability of the Squashing function, it is possible to solve classification problems, where other commonly used activation functions fail.
One-dimensional objects as nanowires have been proven to be building blocks in novel applications due to their unique functionalities. In the realm of magnetic materials, iron-oxides form an important class by providing potential solutions in catalysis, magnetic devices, drug delivery, or in the field of sensors. The accurate composition and spatial structure analysis are crucial to describe the mechanical aspects and optimize strategies for the design of multi-component NWs. Atom probe tomography offers a unique analytic characterization tool to map the (re-)distribution of the constituents leading to a deeper insight into NW growth, thermally-assisted kinetics, and related mechanisms. As NW-based devices critically rely on the mechanical properties of NWs, the appropriate mechanical modeling with the resulting material constants is also highly demanded and can open novel ways to potential applications. Here, we report a compositional and structural study of quasi-ceramic one-dimensional objects: α-Fe ⊕ α-FeOOH(goethite) ⊕ Pt and α-Fe ⊕ α-Fe3O4(magnetite) ⊕ Pt core–shell NWs. We provide a theoretical model for the elastic behavior with terms accounting for the geometrical and mechanical nonlinearity, prior and subsequent to thermal treatment. The as-deposited system with a homogeneous distribution of the constituents demonstrates strikingly different structural and elastic features than that of after annealing, as observed by applying atom probe tomography, energy-dispersive spectroscopy, analytic electron microscopy, and a micromanipulator nanoprobe system. During annealing at a temperature of 350 °C for 20 h, (i) compositional partitioning between phases (α-Fe, α-Fe3O4 and in a minority of α-Fe2O3) in diffusional solid–solid phase transformations takes place, (ii) a distinct newly-formed shell formation develops, (iii) the degree of crystallinity increases and (iv) nanosized precipitation of evolving phases is detected leading to a considerable change in the description of the elastic material properties. The as-deposited nanowires already exhibit a significantly large maximum strain (1–8%) and stress (3–13 GPa) in moderately large bending tests, which become even more enhanced after the annealing treatment resulting at a maximum of about 2.5–10.5% and 6–18 GPa, respectively. As a constitutive parameter, the strain-dependent stretch modulus undoubtedly represents changes in the material properties as the deformation progresses.
The theories of multi-criteria decision-making (MCDM) and fuzzy logic both aim to model human thinking. In MCDM, aggregation processes and preference modeling play the central role. This paper suggests a consistent framework for modeling human thinking by using the tools of both fields: fuzzy logical operators as well as aggregation and preference operators. In this framework, aggregation, preference, and the logical operators are described by the same unary generator function. Similarly to the implication being defined as a composition of the disjunction and the negation operator, preference operators were introduced as a composition of the aggregative operator and the negation operator. After a profound examination of the main properties of the preference operator, our main goal is the implementation into neural networks. We show how preference can be modeled by a perceptron, and illustrate the results in practical neural applications.
Most multi-layer neural networks used in deep learning utilize rectified linear neurons. In our previous papers, we showed that if we want to use the exact same activation function for all the neurons, then the rectified linear function is indeed a reasonable choice. However, preliminary analysis shows that for some applications, it is more advantageous to use different activation functions for different neurons – i.e., select a family of activation functions instead, and select the parameters of activation functions of different neurons during training. Specifically, this was shown for a special family of squashing functions that contain rectified linear neurons as a particular case. In this paper, we explain the empirical success of squashing functions by showing that the formulas describing this family follow from natural symmetry requirements.
Interpretable neural networks based on continuous-valued logic and multicriteria decision operators
(2020)
Combining neural networks with continuous logic and multicriteria decision-making tools can reduce the black-box nature of neural models. In this study, we show that nilpotent logical systems offer an appropriate mathematical framework for hybridization of continuous nilpotent logic and neural models, helping to improve the interpretability and safety of machine learning. In our concept, perceptrons model soft inequalities; namely membership functions and continuous logical operators. We design the network architecture before training, using continuous logical operators and multicriteria decision tools with given weights working in the hidden layers. Designing the structure appropriately leads to a drastic reduction in the number of parameters to be learned. The theoretical basis offers a straightforward choice of activation functions (the cutting function or its differentiable approximation, the squashing function), and also suggests an explanation to the great success of the rectified linear unit (ReLU). In this study, we focus on the architecture of a hybrid model and introduce the building blocks for future applications in deep neural networks.
Comparing multidimensional sensor data from vehicle fleets with methods of sequential data mining
(2020)
Reading and understanding large amounts of sensor data from vehicle test drives becomes more and more important. In order to test vehicle components or analyze exhaust emissions in real test drives, the sensor data obtained from these test drives have to be comparable. Otherwise components or exhaust emissions are tested and analyzed under false conditions. The sensor data obtained during test drives are highly multidimensional which makes it even more complicated to identify recurring patterns. We present a process model to compare different test drives according to their sensor data and so give an answer to the question whether or not test drives in different cities, locations and environments are representative to real driving scenarios. The algorithms we use focus on segmentation of the individual multivariate test drive data and on clustering of the segments according to different methods. We present several segmentation and cluster methods and compare which of them is best suited for comparing test drives. The segmentation method we identified as best suited is based on principal component analysis. As cluster methods we examine hierarchical, partitioning and density-based clustering in detail.
This paper uses several techniques to monitor the ageing of commercial LiFePO4 cells, which are cycled at 55 °C and −20 °C at various depths of discharge. Ageing at lower depth of discharge leads to higher capacity fading, as compared to higher depth of discharge. The highest capacity fading is observed using 50% depth of discharge for cycling at 55 °C, while the lowest capacity fading is observed for the cells aged at 100% depth of discharge when cycled at −20 °C. Using incremental capacity analysis and differential voltage analysis the capacity fading is monitored and underlying ageing mechanisms are described. The loss of lithium inventory and the loss of active material, especially on the cathode side, are the major degradation mechanisms for the cells. The first incremental capacity analysis peak of the discharge process can be used in our case to predict remaining life and cell capacity.
The paper presents post-mortem analysis of commercial LiFePO4 battery cells, which are aged at 55 °C and − 20 °C using dynamic current profiles and different depth of discharges (DOD). Post-mortem analysis focuses on the structure of the electrodes using atomic force microscopy (AFM) and scanning electron microscopy (SEM) and the chemical composition changes using energy dispersive X-ray spectroscopy (SEM-EDX) and X-ray photoelectron spectroscopy (XPS). The results show that ageing at lower DOD results in higher capacity fading compared to higher DOD cycling. The anode surface aged at 55 °C forms a dense cover on the graphite flakes, while at the anode surface aged at − 20 °C lithium plating and LiF crystals are observed. As expected, Fe dissolution from the cathode and deposition on the anode are observed for the ageing performed at 55 °C, while Fe dissolution and deposition are not observed at − 20 °C. Using atomic force microscopy (AFM), the surface conductivity is examined, which shows only minor degradation for the cathodes aged at − 20 °C. The cathodes aged at 55 °C exhibit micrometer size agglomerates of nanometer particles on the cathode surface. The results indicate that cycling at higher SOC ranges is more detrimental and low temperature cycling mainly affects the anode by the formation of plated Li.
Drug-induced liver toxicity is one of the most common reasons for the failure of drugs in clinical trials and frequent withdrawal from the market. Reasons for such failures include the low predictive power of in vivo studies, that is mainly caused by metabolic differences between humans and animals, and intraspecific variances. In addition to factors such as age and genetic background, changes in drug metabolism can also be caused by disease-related changes in the liver. Such metabolic changes have also been observed in clinical settings, for example, in association with a change in liver stiffness, a major characteristic of an altered fibrotic liver. For mimicking these changes in an in vitro model, this study aimed to develop scaffolds that represent the rigidity of healthy and fibrotic liver tissue. We observed that liver cells plated on scaffolds representing the stiffness of healthy livers showed a higher metabolic activity compared to cells plated on stiffer scaffolds. Additionally, we detected a positive effect of a scaffold pre-coated with fetal calf serum (FCS)-containing media. This pre-incubation resulted in increased cell adherence during cell seeding onto the scaffolds. In summary, we developed a scaffold-based 3D model that mimics liver stiffness-dependent changes in drug metabolism that may more easily predict drug interaction in diseased livers.
Electrochemical strain microscopy (ESM) is a powerful atomic force microscopy (AFM) mode for the investigation of ion dynamics and activities in energy storage materials. Here we compare the changes in commercial LiFePO4 cathodes due to ageing and its influence on the measured ESM signal. Additionally, the ESM signal dynamics are analysed to generate characteristic time constants of the diffusion process, induced by a dc-voltage pulse, which changes the ionic concentration in the material volume under the AFM tip. The ageing of the cathode is found to be governed by a decrease of the electrochemical activity and the loss of available lithium for cycling, which can be stored in the cathode.