A special feature of this text is its combination of theoretical depth with extensive practice-oriented material, including many exercises, Python-based projects, and real-world case studies that bridge mathematical analysis with implementation and experimentation. Beyond just standard architectures, the book introduces original coalitional neural models with energy-based foundations, drawing on statistical physics, game theory, and random matrix theory to analyze and redesign deep networks at a fundamental level. It concludes with dedicated chapters on the ethical and social implications of large-scale models and on emerging research directions such as topological datat analysis, meta-reasoning in LLMs, and causal inference: helping readers connect core techniques to current debates and future developments in AI. 

Meant for advanced undergraduates, graduate students, researchers, and professionals, this single-author monograph provides a coherent and pedagogically structured treatment suitable for classroom adoption, self-study, and reference. Readers are equipped not only to understand existing models, but also to engage with ongoing research on interpretability, robustness, and the next generation of learning architectures.