Deepen your knowledge of Machine learning with deep learning modeling techniques.

Duration: 3 days

Prerequisites: Basics in python programming, Familiarity with ML concepts

Objectives: Hands-on deep learning techniques

Content

1. Introduction
   • Basic concepts: (Fully connected neural networks, layers, forward propagation)
   • Nonlinear activation functions
   • Loss functions for supervised learning
   • Initialization and regularization
   • Backward propagation
   • Optimization and learning algorithms
   • Introduction to Pytorch
   • Use case: Optical Character Recognition (OCR)
2. Convolutional Neural Networks (CNN) for image recognition
   • Motivation and key-concepts (local connectivity, weight sharing)
   • Convolution: kernels, filters and feature maps
   • Aggregation, downsampling and pooling
   • Popular architectures: VGG, ResNet, GoogleNet
   • Image pre-processing
   • Using a pre trained neural network
3. Recurrent Neural Networks (RNN)
   • Motivation and key-concepts (windows size, memory, etc)
   • RNN most used architectures (LSTM, GRU), their losses and gradients
   • Discrete sequence: One hot encoding and embeddings
   • Application to classification
   • Application to sequence prediction
4. Representation learning
   • Autoencoders
   • Mutual information neural estimator
   • Deep Info Max
   • Contrastive Predictive Coding
5. Generative Models
   • Variational Autoencoders
   • GANs
   • Generation by reinforcement
   • Generation under condition
   • Style transfer
   • Evaluation of generative models
6. Advanced topics
   • Multi-task learning
   • Semi-supervised learning
   • Transfer learning
   • Debugging neural networks