## Practical Machine Learning Course (2019): This page is maintained by Dr Adrian Bevan

This web page contains a set of slides and example scripts used for a Practical Machine
Learning (PML) summer school course delivered at QMUL in 2019.
These resources are prepared using several Python
packages including NumPy, MatplotLib and Tensorflow. As the TensorFlow versions change
from time to time, and this is beyond my control you may find different versions of the
slides and example scripts to match given versions of that code.

**Syllabus:** This course will cover the following technical aspects: introductory python including use of NumPy arrays and plotting functionality, the use of TensorFlow for computation, including multilayer perceptrons and deep networks (including convolutional neural networks). The following concepts will be discussed: linear discriminants, perceptrons, activation functions (binary threshold, logistic, hyperbolic tangent, relu), neural networks, multilayer perceptrons, convolutional neural networks, training and validation for supervised learning problems, dropout, maxpooling, optimisation; function approximation, classification and regression.

**Requirements:**This has been written for

- anaconda/Python 3.6
- TensorFlow 1.13.1

### Ancillary notes

- Linear algebra and calculus: (pdf)
- Common mistakes: (pdf)
- Portfolio submission: (pdf)

### Slides - to be updated

- Introduction : (pdf)
- Python Coding : (pdf)
- TensorFlow Coding : (pdf)
- Linear Regression : (pdf)
- Introductory NNs : (pdf)
- Classification : (pdf)
- Regression : (pdf)
- More TensorFlow : (pdf)
- Function Approximation : (pdf)
- Optimisation : (pdf)
- Deep Learning: MLPs : (pdf)
- Convolutional Neural Networks : (pdf)
- CNNs - TensorFlow : (pdf)
- Higgs to tautau Kaggle Challenge: (pdf)
- Resources : (pdf)
- Ethics : (pdf)

### Example Code - to be updated

- Python Language Examples: (zip) (tgz)
- TensorFlow API Examples: (zip) (tgz)
- TensorFlow Model Examples: (zip) (tgz) [includes Higgs Kaggle example]

### Example Data

- Kaggle Higgs Data Challenge Data (zip)
The data can also be obtained from the Kaggle Higgs Challenge web page, but here I have processed the data into signal, background for each of the four types of example used [train, private test, public test, unused]. Note that the data zip file is 189Mb.

### Guest Lectures

- Prof R. Ashcroft: The Ethics of AI
- Prof M. Barnes: AI in health
- Dr N. Barlow: Data science at the Alan Turing Institute

Back to the *home-page*