Теория глубокого обучения | SOTAML School 2020

Семен Поляков, программист-разработчик команды CoreML ВКонтакте План лекции: Часть 1: 1. Intro • ML vs DL • DL history 2. Architectures • Feedforward networks • CNN • RNN • Transformers Часть 2: 1. Bias and Variance tradeoff • Problem setup • Regularization 2. Who was after Adam? • Optimizers 3. Multi-tasking • Multi-tasking • Transfer learning 4. Compression 5. Last but not least • Adversarial attacks • Neural tangent kernels • Flows • Final