Led by Ajit Jaokar and supported by a number of other AI and coding specialists, the course is based on the book Foundations
- Introduction to machine learning, deep learning and AI
- Introduction to neural networks, TensorFlow and Keras
- Important changes in TensorFlow 2.0
- Introduction to neural networks
- Introduction to Artificial Intelligence applications
Classification – our first example of TensorFlow 2.0
- Multi-layer perceptron (MLP) – our first example of a network
- Problems in training the perceptron and their solutions
- Activation functions – sigmoid, tanh, ReLU and others
Improving the baseline
- Establishing a baseline
- Improving the baseline using various strategies such as
- hidden layers
- dropout
- optimisers
- epochs
- controlling the optimizer learning rate
- increasing the number of internal hidden neurons,
- increasing the size of batch computation, etc
- Regularization to avoid overfitting
Regression
- What is regression?
- Prediction using linear regression
- Simple linear regression, multiple linear regression and multivariate linear regression
- Predicting house price using linear regression
- Logistic regression
- Logistic regression on the MNIST (Modified National Institute of Standards and Technology) dataset
Convolutional Neural Networks
- Deep Convolutional Neural Network (DCNN)
- Local receptive fields
- Shared weights and bias
- Pooling layers, max pooling, average pooling
- LeNet code in TensorFlow 2.0
Natural Language Processing
- Word embedding
Recurrent Neural Networks
- The basic RNN cell
- Backpropagation through time (BPTT)
- Vanishing and exploding gradients
- Long short-term memory (LSTM)
- Gated recurrent unit (GRU)
Autoencoders
- Introduction to autoencoders
- Vanilla autoencoders
- Sparse autoencoder
- Denoising autoencoders
- Clearing images using a denoising autoencoder
- Stacked autoencoder
- Convolutional autoencoder for removing noise from images
Unsupervised Learning
- Principal component analysis
- PCA on the MNIST dataset
- K-means clustering
- Restricted Boltzmann machines
- Reconstructing images using RBM
- Deep belief networks
- Variational Autoencoders
TensorFlow and Cloud
- Deep Learning on Cloud
- Microsoft Azure, Amazon Web Services (AWS), Google Cloud Platform (GCP)
- Virtual machines on Cloud
- TensorFlow Extended for production
Working with real life applications – processes
- Working with real life applications – added Agile, Cloud, etc.
The course will remain open, with the materials still available to you for six additional weeks after the end of the course.
Coding Exercises:
Hands-on coding exercises will be set during the week via Microsoft Teams, with full support provided by the tutors. These will be a mix of individual and group projects.
The coding exercises will relate to the areas discussed in the tutorials:
- Classification – our first example of TensorFlow 2.0: Recognising handwritten digits as a classification example
- Improving the baseline: regularization
- Regression: logistic regression
- Convolutional Neural Networks: CIFAR-10
- Recognizing CIFAR-10 images with deep learning
- Improving the CIFAR-10 performance with a deeper network
- Improving the CIFAR-10 performance with data augmentation
- Predicting with CIFAR-10
- Natural Language Processing
- Using word embeddings for spam detection
- Using BERT (Bidirectional Encoder Representations from Transformers)
- Recurrent Neural Networks: Sentiment Analysis
- Autoencoders: Reconstructing handwritten digits using an autoencoder
- Unsupervised Learning: PCA and K-means