Workshop Machine Learning in TensorFlow/Keras​ Fundamentals

Introduction

• What is ML?
• Where can I find it in real life?
• Why now?
• What are the three main categories of ML?
  – Supervised learning
  – Unsupervised learning
  – Reinforcement learning (demo)
• ML pipeline

Machine Learning with sci-kit

• ML pipeline review
• Scikit Python Library
• Data representation
  • Feature matrix
  • Target array
  • Iris dataset example
• Estimator API
• Linear Regression
  • Simple Linear Regression
  • Model Evaluation
  • Polynomial Regression
• Selecting the best model
• The bias-variance trade-off
• Logistic Regression
  • Who survives the Titanic?
• Naive Bayes
  • Gaussian Naive Bayes
  • Multinomial Naive Bayes
  • Categorical Naive Bayes
• k Nearest Neighbours
• k-Means Clustering
• Dimensionality reduction
  • Principal Components Analysis (PCA)
  • Singular Value Decomposition (SVD)
• Decision Trees
• Random Forests

Hands-on Lab:

• Playing around with different values affecting the bias and the variance, calculating precision, recall, F1 and F2-scores, comparing different models on the training and testing accuracies
• Doing a little bit of data preprocessing, analyzing the difference between categorical and numerical data, plotting some relevant statistical values and visually inspecting the correlation between features

Neural Networks in Tensorflow/Keras

• Artificial Neural Networks (ANNs)
  – Neurons
  – Layers
  – Activation Functions
  – More vocabulary
• Popular Frameworks
• Keras
• Linear Regression
  – Defining Models in Keras
  – Training and predicting
• Fashion MNIST example

Hands-on Lab:

• Creating our first custom neural network model
• Choosing the number of layers and the number of neurons per layer
• Tweaking the learning rate
• Training the neural network on real world data

Convolutional Neural Networks (peek)

• Motivation behind CNNs
• CNN Building blocks
  – Convolution Layers
  – Pooling Layers
• CNNs in Keras
• Data Augmentation
• Architectures

NLP using Deep Learning

• Spam detector
• Sentiment analyzer
• Autocomplete

Reinforcement Learning

• Frozen Lake demo
• Flappy Bird demo

Recommender Systems

• Data preparation
• Cosine distance
• SVD for recommender systems
• Autoencoder demo