Learning Through Research: the Impact of Pattern Extraction on Neural Networks Architectures

Learning is a complex process to generalize. Generations exhibit diverse learning styles across the years. Even individuals, each have his/her own learning style. Majority of today's learners acquire knowledge and understanding, skills and other domains of learning through the use of smart technologies and devices. These new paradigms facilitate and enable global learning environment. Thus communication paradigms shift toward global horizons, setting the need to develop new learning strategies that attract attention and motivate learning. Learning through research is proposed in this paper. The approach is demonstrated via a research on pattern recognition. The focus is on developing a learning strategy through research on the impact of feature extraction on neural networks architectures. Pattern extraction is a crucial stage in any recognition process. This stage affects the complexity of the neural networks in many application fields including pattern recognition. The choice of good features for a neural architecture has great impact on the recognition as well as on the neural network complexity. This in turn will affect the recognition process as a whole. Good features contribute to efficient use of the memory and reduce the number of the network processors within the architecture. This paper proposes a learning strategy that reflects the impact of pattern extraction on neural networks architectures in images recognition systems. Neural networks were developed for image recognition. Different architectures were developed to investigate the impact and contribution of features on the neural network architecture. A visual tool, Feature Extraction and Image Recognition (FEIR) developed to enable insight the system dynamics This strategy provides the learners with a mechanism to investigate and explore..