ME324: Artificial Intelligence and Deep Learning

Artificial intelligence (AI) and deep learning are shaping the world around us, they’re driving innovation and transforming how organisations conduct research, make decisions, and build products. Yet for many, these systems remain a mystery, limiting their ability to adapt, critique, or responsibly deploy them.

This course will equip you with the foundational knowledge necessary to understand how modern AI systems work, with a particular focus on deep neural networks. You will move from first principles through to exploring core training algorithms that power today’s models. 

We will construct deep networks, image classification models, and generative models such as autoencoders. We will also consider how transformer architectures enable large language models such as ChatGPT.

This course takes a hands-on, practical approach. In labs, you will use Python and PyTorch to guide you through building models from scratch, diagnose their behaviour, and improve their performance. Examples and applications will be drawn from social science, policy, and industry, helping you to connect modelling choices to substantive questions and real-world data. We will also discuss the ethical implications of these models.

By the end of the course, you will be able to design, train, and critically evaluate deep learning models for a range of predictive and generative tasks. Rather than relying on off-the-shelf tools, you will gain the confidence to customise, critique, and extend AI systems for your own research or professional projects.

Prerequisites: Students should have a good understanding of general machine learning principles, e.g. by taking ME314 or ME315 prior to this course (or courses at a home institution of a similar level). The course will also rely on familiarity with simple calculus (i.e. basic differentiation). 

The course will use the python programming language, so some exposure to computer programming with this language is required.

Students will be expected to have a laptop that they can bring to class, which can run the required statistical software and do the required data analysis—for example, a recent MacBook Pro/Air or Windows/Linux laptop with at least 8 GB RAM, and not a Chromebook or a tablet. Please contact the instructors if you have any questions or concerns.

Level: 300 level. Read more information on levels in our FAQs

Fees: Please see Fees and payments

Lectures: 36 hours

Classes: 18 hours

Assessment: One take-home assessment and one in-person exam

Typical credit: 3-4 credits (US) 7.5 ECTS points (EU)

Please note: Assessment is optional but may be required for credit by your home institution. Your home institution will be able to advise how you can meet their credit requirements. For more information on exams and credit, read Teaching and assessment

This course is suited to those who want to move beyond “using AI tools” to understanding and building them. It will appeal to students in data science, economics, political and other social sciences, public policy, and business who are comfortable with basic Python and quantitative reasoning. The skills developed are relevant for careers in data science and analytics, AI and machine learning, tech and management consulting, product management, policy analysis, and research roles across industry, government, and NGOs.

• Understand core concepts underpinning modern deep learning, including computational graphs, backpropagation, and key neural network architectures.
• Design, implement, and train deep learning models in Python/PyTorch.
• Evaluate and diagnose model performance using appropriate metrics, visualisations, and experiments.
• Critically assess the opportunities, limitations, and ethical implications of applying deep learning in real-world research, policy, and industry settings.

The design of this course is guided by LSE faculty, as well as industry experts, who will share their experience and in-depth knowledge with you throughout the course.