Mathematical Neural Networks

ANNs succeed in several tasks for real scenarios due to their high learning abilities. This paper focuses on theoretical aspects of enhance the capacity implementing those modifications that make absorb defining features each scenario. work may be also encompassed within trend devoted providing mathematical explanations ANN performance, with special attention activation functions. The base algorithm has been mathematically decoded analyse required functions regarding impact training process and applicability Universal Approximation Theorem. Particularly, significant new results identify which undergo some usual failings (gradient preserving) are presented here. is first paper—to best author’s knowledge—that stresses role injectivity functions, received scant literature but great incidence performance. In this line, a characterization injective provided related monotonic satisfy classical contractive condition as particular case Lipschitz A summary table these provided, targeted at documenting how select function situation.