Media Summary: Unfortunately, the first lecture did not get recorder. This is an old recording. In this lecture, we go through the idea of data-driven ... In this lecture, we go through self-attention mechanism and transformer architecture. In this lecture we get into details of components used in CNNs.
Uoft Ece1508 Applied Deep Learning - Detailed Analysis & Overview
Unfortunately, the first lecture did not get recorder. This is an old recording. In this lecture, we go through the idea of data-driven ... In this lecture, we go through self-attention mechanism and transformer architecture. In this lecture we get into details of components used in CNNs. In this lecture we go through Auto-Encoders, studying vanilla AE, sparse AE, denoising AE and finally variational AE that can be ... We unfold the problem of overfitting, try to develop a solution called Regularization and then get to Dropout idea. In this lecture we go through Seq2Seq models and build a simple language model.
In this lecture, we complete our discussions on Gradient Descent. We then start with Fully Connected FNNs and discuss how we ... In this lecture, we go through batch normalization idea. We then start chapter 4, where we introduce convolutional neural networks ... In this lecture, we discuss the idea of data augmentation, synthetic data generation and data cleaning. We understand the concept ... This lecture goes through ResNet and the idea of skip connection.
