FIT4009 Advanced topics in intelligent systems - Semester 2 , 2007

Unit leader :

David Albrecht

Lecturer(s) :

Clayton

  • David Albrecht
  • Ann Nicholson
  • Geoff Webb
  • David Dowe
  • David Green

Introduction

Welcome to FIT4009 Advanced Topics in Intelligent Systems.

Intelligent Systems is one of the most active research groups in the Faculty of IT. In 2007, this unit will be divided into topics, each introducing a current research area. Each topic will involve 4 hours of lectures over 2 weeks.

  • Topic 1: Introduction to Bayesian Networks (Ann Nicholson)

    Bayesian networks have rapidly become one of the leading technologies for reasoning with uncertainty and applying AI to real world problems. This follows the work of Pearl, Lauritzen, and others in the late 1980s showing that Bayesian reasoning in practice could be tractable (although in principle it is NP-hard). We begin this topic with a brief examination of the philosophy of Bayesianism, including a quick review of probability theory. We then introduce Bayesian networks. We then cover the syntax and semantics of BN and how to reason with them. The BN software package Netica will be introduced and used in this topic.

  • Topic 2: Decision Analysis (David Albrecht)

    In a wide variety of areas, including medical diagnosis, business investments, oil exploration, and weather prediction, people develop models to assist them in making rational decisions. This topic will provide an introduction to Decision networks and how they can be used in the decision making process.

    We begin with an introduction to decision analysis, motivating the use of performance measures (utilities) in decision making. We then describe how Bayesian networks can be extended to Decision networks, to handle: dynamic domains, choices of actions, and utilities.

  • Topic 3: Complex Systems (David Green)

    Complexity is the richness in structure and behaviour often seen in large systems. It is closely associated the emergence of global features from local interactions, as captured in the popular saying "the whole is greater than the sum of its parts." For example, a flock of birds emerges when individual birds coordinate their behaviour with each other. Related fields of research include: network theory, non-linear systems and chaos, Artificial Life and Evolutionary Computing.

  • Topic 4: Learning Classification Models (Geoff Webb)

    Rapid growth in data storage and accessibility has fuelled demand for computational techniques that extract information from data. Classification learning forms models from data that can be used for categorisation. These techniques are central to the operations of Google and other internet corporations, have many applications in commerce, and play an increasingly significant role in modern science. This topic will introduce the area of classification learning, presenting some key techniques and how to apply them.

  • Topic 5: Minimum Message Length (MML) (David Dowe)

    • Minimum Message Length (MML) is a universal and unifying way of doing machine learning, statistics, econometrics, inductive inference and (what is fashionably known by many as) ``data mining''. Given data, D, MML seeks the hypothesis (or theory), H, which gives the shortest two-part encoding:

    [encoding of H] [encoding of D given H].

    MML was developed by Chris Wallace and David Boulton in 1968. It is a generalisation of Ockham's razor and works very well as such - despite (ongoing?) misunderstandings (or confusions?) of some authors on the topic of Ockham's razor. As one person once described such a misunderstanding: ``... usual confusion between inference and prediction''.

    The relationship between MML and Kolmogorov complexity ensures that, at least in principle, MML can infer any underlying computable function (with or without noise) in a body of data - in this sense, it is universal. In practice, the mathematics gets challenging or just too CPU-intensive and we often resort to approximations.

    MML has the Midas touch of giving state-of-the-art best practice solutions to a range of inference problems, and it also does this with its Ockham-like preference for simplicity.

    Unit synopsis

    Methods from Artificial Intelligence (AI) form the basis for many advanced information systems. These techniques address problems that are difficult to solve or not efficiently solvable with conventional techniques. Building on the undergraduate curriculum this unit introduces the student to advanced AI methods and their applications in information systems.

    Learning outcomes

    Knowledge and Understanding

    With successful completion of the unit the you

    • will have achieved an overview of different technologies that form the basis of intelligent information systems,
    • will have understood the capabilities of these methods,
    • will have learned to recognize tasks that can be solved with these methods,
    • will be able to judge the limitations of these methods.

    Attitudes, Values and Beliefs

    With successful completion of the unit you

    • will be able to apply the standard techniques in the chosen sub-fields of intelligent information systems to the construction and design of such systems
    • will be able to critically evaluate the performance of these approaches
    • will be able to compare these techniques to alternative approaches,
    • will have gained an appreciation of the practical relevance of intelligent information systems.

    Workload

    For on campus students, workload commitments are:

    • Lectures: 3 hours per week
    • Reading, preparation and assignment work: 9 hours per week

    Unit relationships

    Prerequisites

    Completion of the Bachelor of Computer Science or equivalent to the entry requirements for the Honours program. Students must also have enrolment approval from the Honours Coordinator

    Relationships

    This unit is an elective in the Honours Degree of the Bachelor of Computer Science.

    Continuous improvement

    Monash is committed to ‘Excellence in education' and strives for the highest possible quality in teaching and learning. To monitor how successful we are in providing quality teaching and learning Monash regularly seeks feedback from students, employers and staff. Two of the formal ways that you are invited to provide feedback are through Unit Evaluations and through Monquest Teaching Evaluations.

    One of the key formal ways students have to provide feedback is through Unit Evaluation Surveys. It is Monash policy for every unit offered to be evaluated each year. Students are strongly encouraged to complete the surveys as they are an important avenue for students to "have their say". The feedback is anonymous and provides the Faculty with evidence of aspects that students are satisfied and areas for improvement.

    Student Evaluations

    The Faculty of IT administers the Unit Evaluation surveys online through the my.monash portal, although for some smaller classes there may be alternative evaluations conducted in class.

    If you wish to view how previous students rated this unit, please go to http://www.monash.edu.au/unit-evaluation-reports/

    Over the past few years the Faculty of Information Technology has made a number of improvements to its courses as a result of unit evaluation feedback. Some of these include systematic analysis and planning of unit improvements, and consistent assignment return guidelines.

    Monquest Teaching Evaluation surveys may be used by some of your academic staff this semester. They are administered by the Centre for Higher Education Quality (CHEQ) and may be completed in class with a facilitator or on-line through the my.monash portal. The data provided to lecturers is completely anonymous. Monquest surveys provide academic staff with evidence of the effectiveness of their teaching and identify areas for improvement. Individual Monquest reports are confidential, however, you can see the summary results of Monquest evaluations for 2006 at http://www.adm.monash.edu.au/cheq/evaluations/monquest/profiles/index.html

    Unit staff - contact details

    Unit leader

    Dr David Albrecht
    Senior Lecturer
    Phone +61 3 990 55526
    Fax +61 3 990 55157

    Lecturer(s) :

    Dr David Albrecht
    Senior Lecturer
    Phone +61 3 990 55526
    Fax +61 3 990 55157
    Associate Professor Ann Nicholson
    Associate Professor
    Phone +61 3 990 55211
    Fax +61 3 990 55146
    Professor Geoff Webb
    Professor
    Phone +61 3 990 53296
    Fax +61 3 990 55157
    Associate Professor David Dowe
    Associate Professor
    Phone +61 3 990 55776
    Fax +61 3 990 55157
    Professor David Green
    Professor
    Phone +61 3 990 53912
    Fax +61 3 990 55146

    Teaching and learning method

    The main teaching forum will be the lectures. Students are also expected to make use of the on-line discussion forums for any questions regarding the unit material or organisation.

    Communication, participation and feedback

    Monash aims to provide a learning environment in which students receive a range of ongoing feedback throughout their studies. You will receive feedback on your work and progress in this unit. This may take the form of group feedback, individual feedback, peer feedback, self-comparison, verbal and written feedback, discussions (on line and in class) as well as more formal feedback related to assignment marks and grades. You are encouraged to draw on a variety of feedback to enhance your learning.

    It is essential that you take action immediately if you realise that you have a problem that is affecting your study. Semesters are short, so we can help you best if you let us know as soon as problems arise. Regardless of whether the problem is related directly to your progress in the unit, if it is likely to interfere with your progress you should discuss it with your lecturer or a Community Service counsellor as soon as possible.

    Unit Schedule

    Week Topic Key dates
    1 Bayesian Networks  
    2 Bayesian Networks  
    3 Decision Analysis August 3: Bayesian Network Homework due
    4 Decision Analysis  
    5 Complex Systems August 17: Decision Analysis Homework due
    6 Complex Systems  
    7 Classification Models August 31: Complex Systems Homework due
    8 Classification Models  
    9 Minimum Message Length September 14: Classification Models Homework due
    10 Minimum Message Length  
    Mid semester break
    11   October 1: Minimum Message Length Homework due
    12    
    13   October 19: Assignments due

    Unit Resources

    Prescribed text(s) and readings

    There are no prescribed books.

    Recommended text(s) and readings

    • Green (2004), The Serendipity Machine, Allen and Unwin, Sydney. 
    • Korb & Nicholson (2004), Bayesian Artificial Intelligence , Chapman Hall / CRC Press.
    • Raiffa (1970), Decision Analysis: Introductory Lectures on Choices under Uncertainty, Addison-Wesley.
    • Russell & Norvig (2003), Artificial Intelligence: A Modern Approach, Prentice Hall. 
    • Wallace (2005), Statistical and inductive inference by Minimum Message Length,Springer.

    Study resources

    Study resources we will provide for your study are:

    Study resources we will provide for your study are:

    • Lecture notes including required readings;
    • Exercises after each topic;
    • Supplementary material;
    • Discussion groups;
    • This Unit Guide outlining the administrative information for the unit;
    • The FIT4009 unit web site on MUSO, where resources outlined above will be made available.

    Library access

    The Monash University Library site contains details about borrowing rights and catalogue searching. To learn more about the library and the various resources available, please go to http://www.lib.monash.edu.au.  Be sure to obtain a copy of the Library Guide, and if necessary, the instructions for remote access from the library website.

    Monash University Studies Online (MUSO)

    All unit and lecture materials are available through the MUSO (Monash University Studies Online) site. You can access this site by going to:

    1. a) https://muso.monash.edu.au or
    2. b) via the portal (http://my.monash.edu.au).

    Click on the Study and enrolment tab, then the MUSO hyperlink.

    In order for your MUSO unit(s) to function correctly, your computer needs to be correctly configured.

    For example :

    • MUSO supported browser
    • Supported Java runtime environment

    For more information, please visit

    http://www.monash.edu.au/muso/support/students/downloadables-student.html

    You can contact the MUSO Support by: Phone: (+61 3) 9903 1268

    For further contact information including operational hours, please visit

    http://www.monash.edu.au/muso/support/students/contact.html

    Further information can be obtained from the MUSO support site:

    http://www.monash.edu.au/muso/support/index.html

    Assessment

    Unit assessment policy

    Homework Exercises (30%)

    Each topic will have an associated homework exercise, expected to take approximately 4 hours work to complete. The nature of these vary according to the topic. Each will be worth 6% of the final mark, giving a total of 5 x 6 = 30%.
    • Topic 1: Bayesian Network Homework
    • Topic 2: Decision analysis Homework
    • Topic 3: Complex Systems Homework
    • Topic 4: Learning Classification Models Homework
    • Topic 5: Minimum Message Length Homework

    Assignments (2) (35% each = 70% total)

    Each student will be asked to choose TWO of the topics covered that they wish to explore further. The lecturer for that topic has set an assignment on that topic (that may include additional reading).
    • Topics 1 & 2: Bayesian Newtworks and Decision Analysis Assignment
    • Topic 3: Complex Systems Assignment
    • Topic 4: Learning Classification Models Assignment
    • Topic 5: MML Assignment

    Assignment tasks

    • Assignment Task
      Title :
      Bayesian Networks Homework
      Description :
      Weighting :
      6%
      Criteria for assessment :
      Will be included with handout.
      Due date :
      August 3
    • Assignment Task
      Title :
      Decision Analysis Homework
      Description :
      Weighting :
      6%
      Criteria for assessment :
      Will be included with handout.
      Due date :
      August 17
    • Assignment Task
      Title :
      Complex Systems Homework
      Description :
      Weighting :
      6%
      Criteria for assessment :
      Will be included with handout.
      Due date :
      August 31
    • Assignment Task
      Title :
      Classification Models Homework
      Description :
      Weighting :
      6%
      Criteria for assessment :
      Will be included in handout.
      Due date :
      September 14
    • Assignment Task
      Title :
      Minimum Message Length Homework
      Description :
      Weighting :
      6%
      Criteria for assessment :
      Will be included in handout.
      Due date :
      October 1
    • Assignment Task
      Title :
      Assignment 1
      Description :

      You must choose one from the following list of assignments:

      • Bayesian Network and Decision Analysis Assignment
      • Complex Systems Assignment
      • Classification Assignment
      • Minimum Message Length Assignment
      Weighting :
      35%
      Criteria for assessment :
      Due date :
      October 19
    • Assignment Task
      Title :
      Assignment 2
      Description :

      You must choose one from the following list of assignments, that is different from your choice for Assignment 1:

      • Bayesian Network and Decision Analysis Assignment
      • Complex Systems Assignment
      • Classification Assignment
      • Minimum Message Length Assignment
      Weighting :
      35%
      Criteria for assessment :
      Due date :
      October 19

    Examinations

    Assignment submission

    All written assignments need to be submitted to the assignment boxes at the General office on the Clayton campus and must have an coversheet. All other components of the assignments must be submit to the MUSO site for this unit.

    Assignment coversheets

    Coversheets can be obtained from:  

    http://www.infotech.monash.edu.au/resources/student/assignments/

    University and Faculty policy on assessment

    Due dates and extensions

    The due dates for the submission of assignments are given in the previous section. Please make every effort to submit work by the due dates. It is your responsibility to structure your study program around assignment deadlines, family, work and other commitments. Factors such as normal work pressures, vacations, etc. are seldom regarded as appropriate reasons for granting extensions. Students are advised to NOT assume that granting of an extension is a matter of course.

    Requests for extensions must be made to the unit coordinator at least two days before the due date. You will be asked to forward original medical certificates in cases of illness, and may be asked to provide other forms of documentation where necessary. A copy of the email or other written communication of an extension must be attached to the assignment submission.

    Late assignment

    Assignments received after the due date will be subject to a penalty of 5% per day, including weekends. Assignments received later than one week (seven days) after the due date will not normally be accepted.

    Return dates

    Students can expect assignments to be returned within two weeks of the submission date or after receipt, whichever is later.

    Assessment for the unit as a whole is in accordance with the provisions of the Monash University Education Policy at:

    http://www.policy.monash.edu/policy-bank/academic/education/assessment/

    Plagiarism, cheating and collusion

    Plagiarism and cheating are regarded as very serious offences. In cases where cheating  has been confirmed, students have been severely penalised, from losing all marks for an assignment, to facing disciplinary action at the Faculty level. While we would wish that all our students adhere to sound ethical conduct and honesty, I will ask you to acquaint yourself with Student Rights and Responsibilities (http://www.infotech.monash.edu.au/about/committees-groups/facboard/policies/studrights.html) and the Faculty regulations that apply to students detected cheating as these will be applied in all detected cases.

    In this University, cheating means seeking to obtain an unfair advantage in any examination or any other written or practical work to be submitted or completed by a student for assessment. It includes the use, or attempted use, of any means to gain an unfair advantage for any assessable work in the unit, where the means is contrary to the instructions for such work. 

    When you submit an individual assessment item, such as a program, a report, an essay, assignment or other piece of work, under your name you are understood to be stating that this is your own work. If a submission is identical with, or similar to, someone else's work, an assumption of cheating may arise. If you are planning on working with another student, it is acceptable to undertake research together, and discuss problems, but it is not acceptable to jointly develop or share solutions unless this is specified by your lecturer. 

    Intentionally providing students with your solutions to assignments is classified as "assisting to cheat" and students who do this may be subject to disciplinary action. You should take reasonable care that your solution is not accidentally or deliberately obtained by other students. For example, do not leave copies of your work in progress on the hard drives of shared computers, and do not show your work to other students. If you believe this may have happened, please be sure to contact your lecturer as soon as possible.

    Cheating also includes taking into an examination any material contrary to the regulations, including any bilingual dictionary, whether or not with the intention of using it to obtain an advantage.

    Plagiarism involves the false representation of another person's ideas, or findings, as your own by either copying material or paraphrasing without citing sources. It is both professional and ethical to reference clearly the ideas and information that you have used from another writer. If the source is not identified, then you have plagiarised work of the other author. Plagiarism is a form of dishonesty that is insulting to the reader and grossly unfair to your student colleagues.

    Register of counselling about plagiarism

    The university requires faculties to keep a simple and confidential register to record counselling to students about plagiarism (e.g. warnings). The register is accessible to Associate Deans Teaching (or nominees) and, where requested, students concerned have access to their own details in the register. The register is to serve as a record of counselling about the nature of plagiarism, not as a record of allegations; and no provision of appeals in relation to the register is necessary or applicable.

    Non-discriminatory language

    The Faculty of Information Technology is committed to the use of non-discriminatory language in all forms of communication. Discriminatory language is that which refers in abusive terms to gender, race, age, sexual orientation, citizenship or nationality, ethnic or language background, physical or mental ability, or political or religious views, or which stereotypes groups in an adverse manner. This is not meant to preclude or inhibit legitimate academic debate on any issue; however, the language used in such debate should be non-discriminatory and sensitive to these matters. It is important to avoid the use of discriminatory language in your communications and written work. The most common form of discriminatory language in academic work tends to be in the area of gender inclusiveness. You are, therefore, requested to check for this and to ensure your work and communications are non-discriminatory in all respects.

    Students with disabilities

    Students with disabilities that may disadvantage them in assessment should seek advice from one of the following before completing assessment tasks and examinations:

    Deferred assessment and special consideration

    Deferred assessment (not to be confused with an extension for submission of an assignment) may be granted in cases of extenuating personal circumstances such as serious personal illness or bereavement. Special consideration in the awarding of grades is also possible in some circumstances. Information and forms for Special Consideration and deferred assessment applications are available at http://www.monash.edu.au/exams/special-consideration.html. Contact the Faculty's Student Services staff at your campus for further information and advice.