Refine
Document Type
- Report (4)
Has Fulltext
- yes (4)
Is part of the Bibliography
- no (4)
Keywords
- Quantenlogik (2)
- Abfragesprache (1)
- Artificial neural networks (1)
- CNN (1)
- CQQL (1)
- Erklärbarkeit (1)
- Informatik (1)
- Künstliche neuronale Netze (1)
- Mathematische Modellierung (1)
- Neuronales Netz (1)
Institute
Gegeben sei für ein binäres Klassifikationsproblem ein künstliches, neuronales Netzwerk ann bestehend aus ReLUKnoten und linearen Schichten (convolution, pooling, fully connected). Das Netzwerk ann sei mit hinreichender Genauigkeit an Hand von Trainingsdaten trainiert. Wir werden zeigen, dass ein solches Netzwerk in verschiedene Partitionen des Eingaberaums zerlegt werden kann, wobei jede Partition eine lineare Abbildung der Eingabewerte auf einen klassifizierenden Ausgabewert repräsentiert. Im Weiteren gehen wir von einem einfachen Netzwerk ann aus, bei dem die Eingangswerte Mintermen von Attributwerten entsprechen. Einfach ist ein Netzwerk, wenn es für eine geringe Anzahl von Attributen trainiert wurde und die Anzahl der ReLU-Knoten ebenfalls gering ist. In der Arbeit wird gezeigt, dass jede lineare Partition durch einen CQQL-Ausdruck beschrieben werden kann. Ein CQQL-Ausdruck lässt sich mit Hilfe von Quantenlogik-inspirierten Entscheidungsbäumen beschreiben.
Mobile cyber-physical systems (MCPSs), such as the International Space Station, are equipped with sensors which produce sensor data streams. Continuous changes like wear and tear influence the system states of a MCPS continually during runtime. Hence, monitoring is necessary to provide reliability and to avoid critical damage. Although, the monitoring process is limited by resource restrictions. Therefore, the focal point of the present paper is on time-efficient multi-class data stream anomaly detection. Our contribution is bifid. First, we use a one-versusall classification model to combine a set of heterogeneous one-class classifiers consecutively. Such a chain of one-class classifiers provides a very flexible structure while the administrative overhead is reasonably low. Second, based on the classifier chain, we introduce classifier pre-selection.
Convolutional neural networks are often successfully used for classification problems. Usually, a huge number of weights need to be learnt by use of training data. However, the learnt weights give no insight how the cnn really works. Thus, a cnn can be seen as a black box solution. In our approach we develop a method to generate a commuting quantum query language (cqql) condition from a sample derived from a given cnn or from training input. CQQL is inspired by quantum logic and its conditions obey the rules of Boolean algebra. The evaluation of a cqql condition provides values from the unit interval [0; 1] and establishes therefore an elegant bridge between logic and a cnn. The underlying assumption is that a condition (a logic expression) gives much more understanding than pure cnn weights. Furthermore, the rich theory of Boolean algebra can be used for manipulating logic expressions. After extracting a cqql condition from a cnn or its training data we can use logic as a way to predict classes alternatively to a cnn.
From Gleason's theorem we know that in principle every probability measure can be expressed by Hermitian operators in a separable Hilbert space and the Born rule as part of a quantum mechanical system. However, that theorem is not constructive. For a given discrete and additive probability measure based on a σ-algebra we construct a quantum system with projectors expressing that probability measure.