Neural networks (Computer science)

Gliomas are an aggressive class of brain tumors that are associated with a better prognosis at a lower grade level. Effective differentiation and classification are imperative for early treatment. MRI scans are a popular medical imaging modality to detect and diagnosis brain tumors due to its capability to non-invasively highlight the tumor region. With the rise of deep learning, researchers have used convolution neural networks for classification purposes in this domain, specifically pre-trained networks to reduce computational costs. However, with various MRI modalities, MRI machines, and poor image scan quality cause different network structures to have different performance metrics. Each pre-trained network is designed with a different structure that allows robust results given specific problem conditions. This thesis aims to cover the gap in the literature to compare the performance of popular pre-trained networks on a controlled dataset that is different than the network trained domain.

Deep Neural Networks have been widely applied in many different applications and achieve significant improvement over classical machine learning techniques. However, training a neural network usually requires large amount of data, which is not guaranteed in some applications such as medical image classification. To address this issue, people propose to implement meta learning and ensemble learning techniques to make deep learning trainers more powerful. This thesis focuses on using deep learning equipped with meta learning and ensemble learning to study specific problems.
We first propose a new deep learning based method for suggestion mining.
The major challenges of suggestion mining include cross domain issue and the issues caused by unstructured and highly imbalanced data structure. To overcome these challenges, we propose to apply Random Multi-model Deep Learning (RMDL) which combines three different deep learning architectures (DNNs, RNNs and CNNs) and automatically selects the optimal hyper parameter to improve the robustness and flexibility of the model. Our experimental results on the SemEval-2019 competition Task 9 data sets demonstrate that our proposed RMDL outperforms most of the existing suggestion mining methods.

Machine learning techniques such as deep neural networks have become an indispensable tool for a wide range of applications such as image classification, speech recognition, and sentiment analysis in text. An activation function is a mathematical equation that determines the output of each neuron in the neural network. In deep learning architectures the choice of activation functions is very important to the network’s performance. Activation functions determine the output of the model, its computational efficiency, and its ability to train and converge after multiple iterations of training epochs. The selection of an activation function is critical to building and training an effective and efficient neural network. In real-world applications of deep neural networks, the activation function is a hyperparameter. We have observed a lack of consensus on how to select a good activation function for a deep neural network, and that a specific function may not be suitable for all domain-specific applications.

Deep learning models have been successfully applied to a variety of machine learning tasks, including image identification, image segmentation, object detection, speaker recognition, natural language processing, bioinformatics and drug discovery, among other things. This dissertation introduces Multi-Model Deep Learning (MMDL), a new ensemble deep learning approach for signal classification and event forecasting. The ultimate goal of the MMDL method is to improve classification and forecasting performances of individual classifiers by fusing results of participating deep learning models. The performance of such an ensemble model, however, depends heavily on the following two design features. Firstly, the diversity of the participating (or base) deep learning models is crucial. If all base deep learning models produce similar classification results, then combining these results will not provide much improvement. Thus, diversity is considered to be a key design feature of any successful MMDL system. Secondly, the selection of a fusion function, namely, a suitable function to integrate the results of all the base models, is important. In short, building an effective MMDL system is a complex and challenging process which requires deep knowledge of the problem context and a well-defined prediction process. The proposed MMDL method utilizes a bank of Convolutional Neural Networks (CNNs) and Stacked AutoEncoders (SAEs). To reduce the design complexity, a randomized generation process is applied to assign values to hyperparameters of base models. To speed up the training process, new feature extraction procedures which captures time-spatial characteristics of input signals are also explored. The effectiveness of the MMDL method is validated in this dissertation study with three real-world case studies. In the first case study, the MMDL model is applied to classify call types of groupers, an important fishery resource in the Caribbean that produces sounds associated with reproductive behaviors during yearly spawning aggregations. In the second case study, the MMDL model is applied to detect upcalls of North Atlantic Right Whales (NARWs), a type of endangered whales. NARWs use upcalls to communicate among themselves. In the third case study, the MMDL model is modified to predict seizure episodes. In all these cases, the proposed MMDL model outperforms existing state-of-the-art methods.

According to a March 2019 publication by the National Highway Transportation Safety Administration(NHTSA), 62% of all police-reported accidents in the United States between 2011 and 2015 could have been prevented or mitigated with the use of five groups of collision avoidance technologies in passenger vehicles: (1) forward collision prevention, (2) lane keeping, (3) blind zone detection, (4) forward pedestrian impact, and (5) backing collision avoidance. These technologies work mostly by reducing or removing the risks involved in a lane change maneuver; yet, the Broward transportation management system does not directly address these risk. Therefore, we are proposing a Machine Learning based approach to real-time accident prediction for Broward I-95 using the C5.1 Decision Tree and the Multi-Layer Perceptron Neural Network to address them. To do this, we design a new measure of volatility, Lane Change Volatility(LCV), which measures the potential for a lane change in a segment of the highway. Our research found that LCV is an important predictor of accidents in an exit zone and when considered in tandem with current system variable, such as lighting conditions, the machine learning classifiers are able to predict accidents in the exit zone with an accuracy rate of over 98%.

Today transportation systems are facing big transitions all over the world. We created fly overs, roads under the ground, bridges over the river and ocean to get efficient access and to increase the road connectivity. Our transportation system is more intelligent than ever. Our traffic signaling system became adaptive. Our vehicles equipped with new gadgets and we developed new tools for more efficient analysis of traffic. Our research relies on existing traffic infrastructure to generate better understanding of traffic. More specifically, this research focused on traffic and UAV cameras to extract information about the traffic. Our first goal was to create an automatic system to count the cars using traffic cameras. To achieve this goal, we implemented Background Subtraction Method (BSM) and OverFeat Framework. BSM compares consecutive frames to detect the moving objects. Because BSM only works for ideal lab conditions, therefor we implemented a Convolutional Neural Network (CNN) based classification algorithm called OverFeat Framework. We created different segments on the road in various lanes to tabulate the number of passing cars. We achieved 96.55% accuracy for car counting irrespective of different visibility conditions of the day and night. Our second goal was to find out traffic density. We implemented two CNN based algorithms: Single Shot Detection (SSD) and MobileNet-SSD for vehicle detection. These algorithms are object detection algorithms. We used traffic cameras to detect vehicles on the roads. We utilized road markers and light pole distances to determine distances on the road. Using the distance and count information we calculated density. SSD is a more resource intense algorithm and it achieved 92.97% accuracy. MobileNet-SSD is a lighter algorithm and it achieved 79.30% accuracy. Finally, from a moving platform we estimated the velocity of multiple vehicles. There are a lot of roads where traffic cameras are not available, also traffic monitoring is necessary for special events. We implemented Faster R-CNN as a detection algorithm and Discriminative Correlation Filter (with Channel and Spatial Reliability Tracking) for tracking. We calculated the speed information from the tracking information in our study. Our framework achieved 96.80% speed accuracy compared to manual observation of speeds.

This Thesis surveys the landscape of Data Augmentation for image datasets. Completing this survey inspired further study into a method of generative modeling known as Generative Adversarial Networks (GANs). A survey on GANs was conducted to understood recent developments and the problems related to training them. Following this survey, four experiments were proposed to test the application of GANs for data augmentation and to contribute to the quality improvement in GAN-generated data. Experimental results demonstrate the effectiveness of GAN-generated data as a pre-training metric. The other experiments discuss important characteristics of GAN models such as the refining of prior information, transferring generative models from large datasets to small data, and automating the design of Deep Neural Networks within the context of the GAN framework. This Thesis will provide readers with a complete introduction to Data Augmentation and Generative Adversarial Networks, as well as insights into the future of these techniques.

Many current application domains of machine learning and arti cial intelligence
involve knowledge discovery from text, such as sentiment analysis, document
ontology, and spam detection. Humans have years of experience and training with
language, enabling them to understand complicated, nuanced text passages with relative
ease. A text classi er attempts to emulate or replicate this knowledge so that
computers can discriminate between concepts encountered in text; however, learning
high-level concepts from text, such as those found in many applications of text classi-
cation, is a challenging task due to the many challenges associated with text mining
and classi cation. Recently, classi ers trained using arti cial neural networks have
been shown to be e ective for a variety of text mining tasks. Convolutional neural
networks have been trained to classify text from character-level input, automatically
learn high-level abstract representations and avoiding the need for human engineered
features.
This dissertation proposes two new techniques for character-level learning,
log(m) character embedding and convolutional window classi cation. Log(m) embedding
is a new character-vector representation for text data that is more compact and memory e cient than previous embedding vectors. Convolutional window classi
cation is a technique for classifying long documents, i.e. documents with lengths
exceeding the input dimension of the neural network. Additionally, we investigate the
performance of convolutional neural networks combined with long short-term memory
networks, explore how document length impacts classi cation performance and
compare performance of neural networks against non-neural network-based learners
in text classi cation tasks.

The gripping action as performed by an average person is developed over their life and
changes over time. The initial learning is based on trial and error and becomes a natural
action which is modified as the physiology of the individual changes. Each grip type is a
personal expression and as the grip changes over time to accommodate physiologically
changes, it can be considered to be a grip-signature. lt is postulated that an ANN can deliver a classification mechanism that is able to make
sense of the varying gripping inputs that are linearly inseparable and uniquely attributed
to user physiology. Succinctly, in this design, the stifnulus is characterized by a voltage
that represents the applied force in a grip. This signature of forces is then used to train an
ANN to recognize the grip that produced the signature, the ANN in turn is used to
successfully classify three unique states of grip-signatures collected from the gripping
action of various individuals as they hold, lift and crush a paper coffee-cup. A comparative study is done for three types of classification: K-Means, Backpropagation
Feedforward Neural Networks and Recurrent Neural Networks, with recommendations
made in selecting more effective classification methods.

For many years people have consider the possibility that brain activity might provide
a new channel for communication between a person's brain and the external world.
Brain Computer Interface allows humans to control electronic devices using only
their thoughts. The goal of this project is to provide the users with a basic control of a
prosthetic arm using the signal acquired by an Electroencephalogram (EEG). The
main objective of the research is to demonstrate and provide a system that allows
individuals to obtain control of the device with very little training and very few
electrodes. The research includes the development of an elaborate signal-processing
algorithm that uses an Artificial Neural Network to determine the intentions of the
user and their translation into commands to operate the prosthetic arm.