Основы обучения нейронных сетей

Уровень 0 · материалов: 5

В кластер входят документы, описывающие внутренние механизмы обучения нейронных сетей и их практическую реализацию для распознавания образов, и не входят документы, посвященные архитектурам глубокого обучения или прикладным бизнес-кейсам ИИ.

Общие признаки: реализация распознавания символов, алгоритмы градиентного спуска и обратного распространения ошибки, практическое внедрение нейронов, минимизация функции стоимости

Группа выше: Основы машинного обучения и нейросетей

Материалы

• #26991 — Introduction to Neural Networks for Beginners — Представитель этого кластера

  Нейросети для чайников. Начало — Paul_Smith

  Смысл: The main idea is to explain the fundamental principle of neural networks—learning from experience—through a simplified, practical implementation of a character recognizer, moving away from dry academic terminology to make the concept accessible to novices.

  An introductory guide that explains the basic logic of neural networks by building a simple letter-recognition program in Delphi.

• #2607 — Neural Network Implementation for Curious Programmers in C#

  Нейронные сети для любопытных программистов (с примером на c#) — hakandr

  Смысл: The main idea is to provide a simplified, hands-on C# implementation of a basic neural network to help programmers transition from theoretical knowledge to practical application, specifically for character recognition.

  A practical guide and C# code implementation of a basic neural network for digit recognition, intended for programmers to learn the fundamentals of AI.

• #2301 — Introduction to Perceptrons and Basic Neural Network Implementation

  Нейросети для чайников. Часть 2 — Перцептрон — Paul_Smith

  Смысл: The main idea is to demystify the internal workings of a basic artificial neuron (perceptron) by implementing a simple image recognition system that learns through weight adjustment based on error feedback.

  A beginner-friendly tutorial that explains the mathematical logic of a perceptron and provides a C# implementation to recognize a digit through a basic weight-adjustment learning algorithm.

• #21595 — Neural Networks and Deep Learning: Handwritten Digit Recognition

  Нейросети и глубокое обучение, глава 1: использование нейросетей для распознавания рукописных цифр — SLY_G

  Смысл: The main idea is to provide a conceptual and practical foundation for understanding how neural networks learn to recognize patterns by minimizing a cost function via gradient descent, using handwritten digit recognition as a primary prototype.

  An introductory guide to neural networks and deep learning that explains the transition from simple perceptrons to deep architectures through the practical lens of MNIST digit recognition.

• #8002 — Introduction to Neural Network Training and Backpropagation

  Нейронные сети для начинающих. Часть 2 — Arnis71

  Смысл: The main idea of the text is to provide a conceptual and mathematical foundation for how neural networks learn, specifically focusing on the role of bias neurons and the mechanics of the backpropagation algorithm and gradient descent.

  A comprehensive beginner's guide explaining bias neurons, gradient descent, and the mathematical process of backpropagation to train neural networks.