Deep neural networks have stood out as an efficient approach to solving complex problems in various fields of knowledge, including artificial intelligence, image processing, and statistical modeling. Based on the concept of complex networks, these computational structures utilize connections between multiple layers of neurons to extract patterns and make inferences from data. This study aims to conduct a statistical analysis of deep neural networks, evaluating their structural and dynamic properties based on mathematical metrics and complex network models. To this end, a Python program was developed to model artificial neural networks, allowing the investigation of the influence of synaptic weight variation and network topology on their statistical characteristics. The results obtained indicate the importance of balancing network depth and parameter configuration to achieve optimal performance.
|
