"Machine learning techniques and visualization in big data"

This PhD dissertation aims to study, analyse and develop concepts and techniques of Machine Learning, Deep Learning and Visualization for Big Data, focussing onlow computational power devices (such as wearable devices). Additionally, techniques such as Convolutional Neural Networks (CNNs) will be applied,both for processing the data and extracting useful information. The objectives of the PhD dissertation are:
i. Collection - Storage - Preprocessing: Optimal techniques, for collecting and storing large volumes of data, will be studied and implemented. Data will be stored either locallyor remotely in a non-relational database.
ii. Design and Implementation of a Deep-Learning Neural Network Model: Models of CNNs and RNNs (Recurrent Neural Networks) will be studied, designed, implemented and evaluated in order to extract the final model of the system so as to achieve:
• More efficient distributed training
• Lower computational cost when extracting new models.
• Feasible installation of the developed model on devices with low computational power.
iii. Data Augmentation Techniques: Data augmentation techniques will be developed to increase the classification of the model in cases ofsmall data sets.
iv. Visualization: Visualization Tools will be implemented for the interpretation of the results (e.g. interactive graphs or dynamic tables).