Addressing the need for AI education and deep learning in particular, this unit is designed to be a successive, advanced unit to the level-5 Deep Learning 5215 unit. While foundations of Deep Learning have been covered in FIT3181 (undergrad) and FIT5215 (post-grad), there is currently no follow-up unit to cover more advanced knowledge in deep learning at Monash. In addition, with the provisional launch of the Master of AI (AI), this unit will be essential to the education of AI. The unit will place an emphasis on modern techniques machine learning, deep learning and AI systems including deep reinforcement learning, deep generative models and how they might be applied in sophisticated AI-based applications in computer vision and language technology. Learning activities will focus on designing deep neural networks and CNN-based systems for image classification, representation learning on different types of structured, unstructured and semi-structured data, deep generative models, cognitive vision/language systems and deep reinforcement learning.

Upon successful completion of this unit students should be able to:

1. Analyse problems and big datasets with a range of deep learning tools
2. Design solutions to a real world problem using advanced learning systems, what is involved in designing such systems and strategy to maintain them
3. Describe and apply a range of advanced tools in cognitive and learning systems such as DNN, CNN RNN, LSTM and deep reinforcement learning to selected advanced AI-based systems such as vision and NLP
4. Develop advanced unsupervised feature learning models and representation learning models
5. Communicate the results of an analysis, experiments and learning systems for both specific and broad audiences.

020119

Deep learning (DL) is one of the most highly sought after skills in AI. It is one of the most important breakthroughs in technology and has become a driving force for AI research and applications. This course will focus on advanced learning and cognitive systems which uses modern knowledge in deep learning, machine learning, computer vision and language technology to build AI applications. It covers the foundation as well as advanced knowledge in deep learning and related disciplines. Deep neural networks, Convolutional networks, RNNs, LSTM, GRU and optimization techniques such as Adam, Dropout, BatchNorm will be covered. It then focuses on modern techniques including deep reinforcement learning, deep generative models and how they might be applied in computer vision and language technology tasks. Learning activities include designing deep neural networks and CNN-based systems for image/video classification, representation learning on different types of structured, unstructured and semi-structured data, deep generative models, cognitive vision/language systems and deep reinforcement learning.

On campus

Examination (2 hours and 10 minutes): 40%; In-semester assessment: 60%

Minimum total expected workload equals 12 hours per week comprising:

• Two hours/week lectures
• Two hours/week laboratories

A minimum of 8 hours per week of personal study for completing lab/tutorial activities, assignments, private study and revision.

FIT

FIT5201, FIT5215