BDA312 - Deep Learning Applications

Credit Points: 15 credit points

Workload: 48 hours

Prerequisite: BDA214 Machine learning

Co-requisite: N/A

Aims & Objectives

This is a core unit out of a total of 24 units in the Bachelor of Data Analytics (BDA). This unit addresses the BDA course learning outcomes and complements other courses in related fields by developing students’ specialised knowledge of deep learning algorithms to address complex problems in data analytics and other similar disciplines. For further course information refer to: http://www.mit.edu.au/study-with-us/programs/Bachelor-DA. This unit is part of the AQF level 7 (BDA) course.

Students will learn to plan, design, implement and evaluate deep learning applications in advanced and diverse range of data analytic problems. They will gain knowledge of deep learning strategies, model selection, training, testing and validation in data analytic applications. Students will gain hand on practical experience in deploying and testing deep learning solutions for industrial applications. 

This unit will cover the following topics:

  • Deep learning architecture
  • Data abstraction and model interpretation in deep learning
  • Convolution neural network
  • Recurrent neural network
  • Deep learning system reliability
  • Deep learning in industry applications

Learning Outcomes

4.1 Course Learning Outcomes
The Course learning outcomes applicable to this unit are listed on the Melbourne Institute of Technology’s website: www.mit.edu.au 

4.2 Unit Learning Outcomes
At the completion of this unit students should be able to:
a. Describe the use of deep learning in data analytic applications;
b. Design deep learning architecture in consideration of data abstraction and layered learning;
c. Apply deep learning in a selected industry application with complex problems;
d. Evaluate deep learning applications in consideration of system reliability and data interpretation;
e. Recommend appropriate deep learning solutions in a complex industry application. 

Weekly Topics

This unit will cover the following content:

Week Topics
1 Introduction to deep learning
2 Data abstraction and layered learning
3 Data interpretation
4 Deep learning frameworks
5 Deep forward neural networks
6 Convolution neural network (CNN)
7 Recurrent and recursive neural networks
8 Advanced deep learning architectures
9 Performance metrics and system reliability in deep learning
10 Deep learning applications 1 (e.g. intrusion detection applications)
11 Deep learning applications 2 (e.g. natural language processing)
12 Deep learning applications 3 (e.g. computer vision)

Assessment

Assessment Task Due Date A B Learning Outcomes Assessed
Formative Assignment 1 Part A
Assignment 1 Part B
Week 3
Week 8
5%
10%
  a-c
Assignment 2 Week 11 25%   d-e
Laboratory participation & submission Week 2 - 11 10%   a-e
Final Examination (2 hours) End of trimester   50% a-e
TOTALS   50% 50%  

Task Type: Type A: unsupervised, Type B: supervised.

Class Participation and Contribution
This unit has class participation and student contribution as an assessment. The assessment task and marking rubric will follow the Guidelines on Assessing Class Participation (https://www.mit.edu.au/about-us/governance/institute-rules-policies-and-plans/policies-procedures-and-guidelines/Guidelines_on_Assessing_Class_Participation). Further details will be provided in the assessment specification on the type of assessment tasks and the marking rubrics.

Presentations (if applicable)
For presentations conducted in class, students are required to wear business attire.

Textbook and Reference Materials

Textbook:

  • Josh Patterson, Adam Gibson, Deep Learning: A Practitioner's Approach, "O'Reilly Media, Inc.", 28 Jul 2017 

References:

  • Ghatak, Abhijit, Deep Learning with R, Springer, Singapore, 2019
  • Charniak, Eugene, Introduction to deep learning, The MIT Press, 2019. 
  • Bengio, Yoshua, Ian Goodfellow, and Aaron Courville, Deep learning, MIT press, 2016.
  • Brownlee, Jason, Deep learning with python: Develop deep learning models on theano and tensorflow using keras, Machine Learning Mastery, 2016.

Adopted Reference Style: IEEE

Graduate Attributes

MIT is committed to ensure the course is current, practical and relevant so that graduates are “work ready” and equipped for life-long learning. In order to accomplish this, the MIT Graduate Attributes identify the required knowledge, skills and attributes that prepare students for the industry.
The level to which Graduate Attributes covered in this unit are as follows:

Ability to communicate Independent and Lifelong Learning Ethics Analytical and Problem Solving Cultural and Global Awareness Team work Specialist knowledge of a field of study

Legend

Levels of attainment Extent covered
The attribute is covered by theory and practice, and addressed by assessed activities in which the students always play an active role, e.g. workshops, lab submissions, assignments, demonstrations, tests, examinations.
The attribute is covered by theory or practice, and addressed by assessed activities in which the students mostly play an active role, e.g. discussions, reading, intepreting documents, tests, examinations.
The attribute is discussed in theory or practice; it is addressed by assessed activities in which the students may play an active role, e.g. lectures and discussions, reading, interpretation, workshops, presentations.
The attribute is presented as a side issue in theory or practice; it is not specifically assessed, but it is addressed by activities such as lectures or tutorials.
The attribute is not considered, there is no theory or practice or activities associated with this attribute.