Volumen 7 (2017): Heft 4 (October 2017)

In this paper, a novel neural network is proposed, which can automatically learn and recall contents from texts, and answer questions about the contents in either a large corpus or a short piece of text. The proposed neural network combines parse trees, semantic networks, and inference models. It contains layers corresponding to sentences, clauses, phrases, words and synonym sets. The neurons in the phrase-layer and the word-layer are labeled with their part-of-speeches and their semantic roles. The proposed neural network is automatically organized to represent the contents in a given text. Its carefully designed structure and algorithms make it able to take advantage of the labels and neurons of synonym sets to build the relationship between the sentences about similar things. The experiments show that the proposed neural network with the labels and the synonym sets has the better performance than the others that do not have the labels or the synonym sets while the other parts and the algorithms are the same. The proposed neural network also shows its ability to tolerate noise, to answer factoid questions, and to solve single-choice questions in an exercise book for non-native English learners in the experiments.

A classification system for the segmentation of driving maneuvers and its validation in autonomous parking using a small-scale vehicle are presented in this work. The classifiers are designed to detect points that are crucial for the path-planning task, thus enabling the implementation of efficient autonomous parking maneuvers. The training data set is generated by simulations using appropriate vehicle-dynamics models and the resulting classifiers are validated with the small-scale autonomous vehicle. To achieve both a high classification performance and a classification system that can be implemented on a microcontroller with limited computational resources, a two-stage design process is applied. In a first step an ensemble classifier, the Random Forest (RF) algorithm, is constructed and based on the RF-kernel a General Radial Basis Function (GRBF) classifier is generated. The GRBF-classifier is integrated into the small-scale autonomous vehicle leading to excellent performance in parallel-, cross- and oblique-parking maneuvers. The work shows that segmentation using classifies and open-loop control are an efficient approach in autonomous driving for the implementation of driving maneuvers.

Hebbian learning rule is well known as a memory storing scheme for associative memory models. This scheme is simple and fast, however, its performance gets decreased when memory patterns are not orthogonal each other. Pseudo-orthogonalization is a decorrelating method for memory patterns which uses XNOR masking between the memory patterns and randomly generated patterns. By a combination of this method and Hebbian learning rule, storage capacity of associative memory concerning non-orthogonal patterns is improved without high computational cost. The memory patterns can also be retrieved based on a simulated annealing method by using an external stimulus pattern. By utilizing complex numbers and quaternions, we can extend the pseudo-orthogonalization for complex-valued and quaternionic Hopfield neural networks. In this paper, the extended pseudo-orthogonalization methods for associative memories based on complex numbers and quaternions are examined from the viewpoint of correlations in memory patterns. We show that the method has stable recall performance on highly correlated memory patterns compared to the conventional real-valued method.

Rule extraction from neural networks is a fervent research topic. In the last 20 years many authors presented a number of techniques showing how to extract symbolic rules from Multi Layer Perceptrons (MLPs). Nevertheless, very few were related to ensembles of neural networks and even less for networks trained by deep learning. On several datasets we performed rule extraction from ensembles of Discretized Interpretable Multi Layer Perceptrons (DIMLP), and DIMLPs trained by deep learning. The results obtained on the Thyroid dataset and the Wisconsin Breast Cancer dataset show that the predictive accuracy of the extracted rules compare very favorably with respect to state of the art results. Finally, in the last classification problem on digit recognition, generated rules from the MNIST dataset can be viewed as discriminatory features in particular digit areas. Qualitatively, with respect to rule complexity in terms of number of generated rules and number of antecedents per rule, deep DIMLPs and DIMLPs trained by arcing give similar results on a binary classification problem involving digits 5 and 8. On the whole MNIST problem we showed that it is possible to determine the feature detectors created by neural networks and also that the complexity of the extracted rulesets can be well balanced between accuracy and interpretability.

Electroencephalograph (EEG) data provide insight into the interconnections and relationships between various cognitive states and their corresponding brain dynamics, by demonstrating dynamic connections between brain regions at different frequency bands. While sensory input tends to stimulate neural activity in different frequency bands, peaceful states of being and self-induced meditation tend to produce activity in the mid-range (Alpha). These studies were conducted with the aim of: (a) testing different equipment in order to assess two (2) different EEG technologies together with their benefits and limitations and (b) having an initial impression of different brain states associated with different experimental modalities and tasks, by analyzing the spatial and temporal power spectrum and applying our movie making methodology to engage in qualitative exploration via the art of encephalography. This study complements our previous study of measuring multichannel EEG brain dynamics using MINDO48 equipment associated with three experimental modalities measured both in the laboratory and the natural environment. Together with Hilbert analysis, we conjecture, the results will provide us with the tools to engage in more complex brain dynamics and mental states, such as Meditation, Mathematical Audio Lectures, Music Induced Meditation, and Mental Arithmetic Exercises. This paper focuses on open eye and closed eye conditions, as well as meditation states in laboratory conditions. We assess similarities and differences between experimental modalities and their associated brain states as well as differences between the different tools for analysis and equipment.