Machine Learning and Artificial Intelligence in Python

If you are in receipt of a UK state benefit, you are a full-time student in the UK or a student on a low income, you may be eligible for a reduction of 50% of tuition fees. Please see the below link for full details:

Concessionary fees for short courses

• Explore the landscape of contemporary machine learning (ML) and deep learning.
• Learn how to use a variety of machine-learning algorithms to extract features from the data using Python libraries.
• Familiarise with the concepts of overfitting and regularisation in ML.
• Gain insights on how to face scaling issues in a 'big data' scenario.

Learning takes place on a weekly schedule. At the start of each weekly unit, students are provided with learning materials on our online platform, including one hour of pre-recorded video, often supplemented by guided readings and educational resources. These learning materials prepare students for a one-hour live webinar with an expert tutor at the end of each weekly unit which they attend in small groups. Webinars are held on Microsoft Teams, and provide the opportunity for students to respond to discussion prompts and ask questions. The blend of weekly learning materials that can be worked through flexibly, together with a live meeting with a tutor and their peers, maximise learning and engagement through interaction in a friendly, supportive environment.

Each week's session will consist of pre-recorded lectures and hands-on programming exercises, class discussions and interactive programming demonstrations by the lecturer.

A the end of the course the students will be able to:

• choose the right ML task and evaluation metric for a given ML problem and select a set of ML models to be trained;
• set up a data pre-processing pipeline for data science and machine learning algorithms;
• use Python machine learning tools (namely scikit-learn, TensorFlow and PyTorch) to build up ML models, train and evaluate them on a test set;
• evaluate whether a model overfits or underfits the data and act accordingly (e.g. opportunely regularise and overfitting model);
• to identify the appropriate and most performant model for a given task and tune appropriately the hyperparameters (parameters that cannot be learned by the model).

You will be set independent formative and summative work for this course. Formative work will be submitted for informal assessment and feedback from your tutor, but has no impact on your final grade. The summative work will be formally assessed as pass or fail.

The Department's short online courses are taught at FHEQ Level 4, i.e. first year undergraduate level. FHEQ level 4 courses require approximately 10 hours study per week, therefore a total of about 100 study hours.

English Language Requirements

We do not insist that applicants hold an English language certification, but warn that they may be at a disadvantage if their language skills are not of a comparable level to those qualifications listed on our website. If you are confident in your proficiency, please feel free to enrol. For more information regarding English language requirements please follow this link: https://www.conted.ox.ac.uk/about/english-language-requirements

Before attending this course, prospective students will know:

• All the requirements and topics covered in the "Python Programming for Data Science - Introduction" course, i.e:
• The fundamentals of linear algebra: what is a matrix and how matrix addition and multiplication are performed
• The following fundamental concepts of statistics: mean, median, variance and standard deviation, interquartile range
• The fundamentals of algebra: real and complex numbers, exponential and logarithm, and trigonometric functions
• How to perform fundamental Python operations such as variable creation, numerical operations on scalar, vectors and matrices, iteration through a collection, manipulation of elements in a collection.
• How to use NumPy and pandas to import a dataset and extract important statistics from it using techniques such as split-apply-combine (for example, finding the mean, median or max of a quantitative variable for each category in a categorical variable) 
• Given a dataset, how to select the appropriate visualisation graph depending on the information to be conveyed, and use the matplotlib library to draw it and add title, captions and figure legends.
• How to create and add state and behaviour to a class in Python
• How to use nltk to preprocess a text and convert it to a numerical representation that can be manipulated by information retrieval algorithms.
• What is, at least conceptually or visually, a derivative and a gradient