Data Science Meets Deep Learning

This course is an intermediate level 5-week intensive training program. In these 5 week, you will be fully exposed to deep learning and data science both online, offline and with your supportive community, and build up your project portfolio to achieve your career dreams.

• 1.5 hours teaching + 1.5 hours hands-on in-class project per week
• In-class Kaggle Competition

The lesson table is given as:

Week 1

• Basic concepts behind NLP and Deep Learning
• Representation learning and TF-IDF Models
• Stochastic gadients descent
• Softmax and cross entropy error

In-class Project

Implement a basic Softmax layer from scratch (project)

Week 2

• Non-linearities behind neural networks
• Feed-forward computation for NN
• Back-propagation for NN
• Overfitting, Activation functions, Regularization

In-class Project

Tutorial on TensorFlow

Week 3

• Count-based Word representation
• Neural network-based word2vec
• Inherent connections between Word2vec and TF-IDF models
• Recurrent Neural Networks and Language Model

In-class Project

Word vectors training model

Week 4

• Language Model
• Gradient Vanishing and Exploding
• Bi-directional Extension
• Cell Extension: LSTM or GRU

In-class Project

LSTM sentiment analysis

Week 5

• A brief review on CNN
• Implement multi-channel CNNs for sentiment analysis
• Design the code structure following a standard tensorflow template which can provide a good practice for our daily job.