The University of New South Wales

Minutes of Meeting

A meeting of the Computer Science and Engineering Education Committee (CSE Education Committee 21/04)
held at 10:00am on Friday, 09 July 2021, via Zoom.

---

Items

1. Apologies and Welcome

   Present:	Members: John Shepherd (jas) (chair), Andrew Taylor (at), Aaron Quigley (aq), Alan Blair (ab), Oliver Diessel (od), Bruno Gaeta (bg), Eric Martin (em), Fethi Rabhi (fr), Wayne Wobcke (ww) Academic Staff: Boualem Benatallah (bb), Alfred Krzywicki (ak), Kaiqi Liang (kl), Gelareh Mohammadi (gm), Carroll Morgan (cm), Sri Parameswaran (sri), Raveen de Silva (rds), Matthew Sladescu (ms), Hayden Smith (hs), Lina Yao (ly)
   In Attendance:	TSAs: Sebastian Kandi (sk), Nicola Kwan (nk) Guests: Christopher Di Bella (cjdb) (scribe), Lesa de Leau (ll) Students: Michael Gribben (mg), Jashank Jeremy (jj), Zac Kologlu (zk), Tom Kunc (tk), Raghav Lall (rl) Nick Patrikeos (np), Shrey Somaya (ss)
   Apologies:	-
   Absent:	-




2. Minutes of Previous Meeting *

   CSE Teaching Committee 21/03 (4 June 2021)

   • Somebody's name was misspelt?



3. Reports from Bodies outside CSE

   Proposals at recent Engineering Academic Programs Committee generated some discussion:

   • MERE4810 Data Analytics and Digital Technologies for MER
   • Humanitarian Engineering minor
   
   

   MERE4810

   jas:

   there seems to have been a sequence of proposals involving ML
   if I suggest "what about us? we teach ML!", the response is "we're not teaching ML, just using it"
   but they invariably spend a couple of weeks "introducing" ML
   they should consult with CSE
   we could argue that we're turning into maths (service discipline)
   maybe a first year ML course for all engineers so they can save time in these courses

   at:

   I thought that but with a COMP1010 prereq

   bruno:

   combine with intro to stats/algos
   some would argue it's fundamental for sci degrees

   jas:

   too much of this would lead to conflict with MATH

   at:

   DATA1001/9001 is gentle on the ML

   COMP1010

   jas:

   how can we advertise it to the rest of the world?
   tried to get SCI to publicise it
   how do you get gened courses known about?
   would be good to get more students into COMP1010
   the original goal was to have 500+ each term

   Humanitarian Engineering

   ww:

   approved, but details still being developed
   students would need to do an ARTS course in second year
   Awkward: better to have in first year
   Need to do a thesis in their course
   Can still do some projects but can't do as much as normal due to requirements
   They need to do one project (thesis); I think it's beneficial for them to do two

   ab:

   is that in place of gened?
   

   ww:

   not sure
   

   jas:

   specialisation?
   

   ww:

   it's approved as a minor
   putting into a degree is hard
   

   jas:

   can probably fit it into SENG degree easier than others
   

   DESN3000

   ww:

   DESN3000 hasn't run yet
   doing regular work on developing it; final shape should be clear by Dec

   Horizons

   aq:

   would you like me to report on the Horizons Development?
   what the future of edu looks like wrt lifelong learning
   e.g. Telstra
   Telstra recognises they have a need for ML, they can ask us to make bespoke ML courses
   rapid building at scale
   assessment of how this works
   key picture:
   - they require upskilling, Horizons will figure out what the requirements are
   - set up a small adivsory board
   - heads of schools would set up the right academics
   - academics update/create new content which can be used in UNSW too
   - deliver to the client
   none of these exist
   there's client orgs and Horizons who sit between them and the unis
   look at what AGSM does, which lists all the short courses on their website
   what is good for Telstra might also be useful for BHP
   nothing underway atm afaik
   sounds similar to SecEdu but not just for CommBank and is open
   they haven't come to me without any proposals
   whatever content we produce could be useful for our undergrad courses so we're not duplicating work
   some schools might make this a part of their perm budget because of lost revenue
   CSE isn't in that position right now
   this is futureproofing in case the nature of undergrad education changes
   one of the most interesting things for CSE is that if a client wanted to send a set of engineers, we can set up a strong set of relationships
   

   jas:

   any ideas on timeframe from request -> product?
   

   aq:

   none yet, but I'm expecting they'll have realistic expectations
   

   jas:

   prior exp suggests it'll take a long time and not be easily integrateable
   




4. New Course: COMP6xxx Recommender Systems

   ly:

   we aim to offer students comprehensive about theoretical and practical knowledge of recommender systems
   how do real-world systems use them?
   good in-depth courage
   based on own research exp in this area
   able to transfer knowledge
   have industrial partners for guest lectures
   since last year:
   we started to design this course
   consulted colleagues to ask for feedback
   coures relevant to COMP9318, and they introduced data mining and ML topics
   small overlap with 9318:
   they only briefly introduce stuff we deeply go into
   COMP6714 covers knowledge graph like us, but they cover how to generate it; we focus on how to use the domain knowledge
   COMP9417 also has some overlap, but we don't come at it from the same angle
   COMP9314/9444 they're more general ML courses; no overlap at all
   

   jas:

   research, industry... ticks all the boxes
   you said it's an ADK course?
   

   ly:

   yes
   

   jas:

   but it's also an intro?
   

   ly:

   COMP9444 is a prereq
   needs a little foundational knowledge
   that's why I designed the foundation part for this course
   

   jas:

   have you looked at COMP4418?
   

   ly:

   no, I didn't check that one in detail, but I scanned all the courses
   I think the more closely related ones are the ones I already mentioned, but they're still different
   

   jas:

   if this is an ADK course, be careful about saying "introductory"
   committee will get upset that it says "ADK" and "introductory"
   

   at:

   expected enrollment numbers? lecturers? industry lecturers?
   

   ly:

   approx 100-150 students
   I will teach if approved, Boualem/Wayne could also teach it
   

   ww:

   AI should really be done before this
   might be difficult for postgrads
   we should think about sequencing
   Deep Learning is offered twice a year; has ~700 students
   I think 100 is a low number!
   

   ly:

   proposed for 22T2 or should it be 22T3?
   

   jas:

   not T3
   

   at:

   what about T1 to facilitate postgrads?
   offered yearly?
   

   ly:

   yes, offered yearly
   

   ss:

   seems to be very practical
   exam is noted to be 60%, why is it so high if the course is really practical?
   

   ly:

   exam is 40%
   

   at:

   propsal says 60%
   

   ss:

   could be some issues because of flex week, which this course doesn't seem to consider
   

   jas:

   assume you're dealing with nine weeks, not ten
   

   ss:

   with the prereq of basic linear algebra
   make it explicit in the handbook because undergrads don't need to do any maths and end up in this course
   probably don't need to make MATH1231 as a prereq
   

   at:

   could get a really heterogeneous set of students despite prereqs
   

   ww:

   what does Alan have to say?
   

   ab:

   for 9444, we assume some knowledge and probability theory
   and linear algebra:
   if students haven't done that, maybe we should create some videos
   

   ww:

   best not to make lin alg a prereq
   

   jj:

   my concern is for postgrads and maths
   

   ab:

   what about UNSW online?
   

   aq:

   I've attended RecSys, which has 50% industry at a 900+ attendance
   I think the popularity could go through the roof
   lists many companies that would benefit from this
   

   jj:

   is there a plan to handle this scaling up?
   

   ly:

   for the first year, I'd like to keep the same scale
   second year and beyond:
   maybe think of some strategies to scale up
   haven't thought about that yet
   consult with colleagues about how to handle this problem, but best to keep the original number for year 1
   

   jas:

   tutors will be an issue for year 1
   

   ly:

   staff or HDR research students
   

   ww:

   have you thought about software tools and assignments
   

   ly:

   yes
   assignments already developed for first offering
   slight change each year
   little-to-no overlap with ML
   

   boualem:

   we might be asking students to apply recommender systems in other course assignments, something to be aware of
   

   jas:

   Are there any objections to moving this forward to the next committee, pending fixes?
   

   Vote Passed




5. New Course: COMP6991 Rust Programming

   at:

   I'd prefer a different name like 'Safe System Programming using Rust'
   

   cjdb:

   Note: use of 'safe' in names upsets safety critical companies in the ISO world
   

   at:

   It's clear Rust is starting gaining prominence
   tutes/labs have focus on "how to rust"
   lectures have theory and concepts
   prereqs: COMP2521 or COMP9024
   ug-only: COMP1521? but no postgrad equivalent
   offered once a year, only if possible
   and this should be explicitly noted
   much bigger emphasis on safe systems; a strategic area for CSE
   

   aq:

   is this covered anywhere else?
   

   zk:

   speak to the fact about what makes this course for UNSW
   we've had a long and rigorous history
   title can be a bit deceiving; maybe "Rigorous/reliable Programming"
   heavily theoretical course, different to COMP6771
   looks at the pitfalls of PLs and how Rust has solved them
   first mainstream language that has taken memory safety
   steals ideas from ML, Haskell
   lots of traction in industry: Firefox, Microsoft, Google
   Rust is particluarly interesting because it achieves it with very little run-time cost
   patches are being adopted to get into Linux kernel
   first major one since C
   students have been increasingly interested in it
   StackOverflow has seen it voted as the most-loved language for six years running
   designed to have a great sense of familiarity with industry programmers
   Rust doesn't have GADTs, which will probably upset Haskell folks
   but it can be picked up by more mainstream programmers
   

   ke:

   on this proposal... a big fan of Rust
   concern is pointing out Rust's solutions to problems... will (UG) students be able to appreciate these complex problems?
   where does this [course] sit?
   we have this problem in OS
   they kind of don't have the background, and I need to help a lot
   

   zk:

   two half-answers
   one:
   what Andrew mentioned earlier; lectures should be very concept-driven
   lecturer demos how things went wrong in other languages
   two:
   as a six level course, students should be enrolled in their third year
   

   ke:

   will students even understand about garbage collection?
   struggling to see where this course fits in
   

   ww:

   reading the ACM's curriculum guidelines
   is it a PL course? an OS course?
   

   zk:

   not an OS course
   

   ke:

   you need a bit of PL, OS, GC, ...
   Rust approaches these in novel ways
   but you don't get the appreciation unless you've got some experience
   

   aq:

   Could this be framed as a “comparative programming languages”
   

   jj:

   what's the point of this course?
   I don't get why we need PL courses (in general)
   a lot of this stuff could be easily integrated into other courses
   

   tk:

   chicken-and-egg problem
   context benefits this course (or whatever you take on later) regardless of when it's taken
   

   ss:

   how do other unis do this?
   Stanford did this as a CS 102 course
   several similar issues were raised, but there's never a great spot for this course
   idea isn't about teaching rust, but rather using rust to improve engineers
   

   jj:

   you just turned this into 3161
   

   zk:

   I disagree
   course has been in the works for 8-10 months, no real good spot for this
   

   jas:

   what about longevity?
   

   zk:

   I don't think it's a problem
   

   ww:

   who will teach?
   

   jas:

   Andrew?
   Zac, anyone else?
   

   zk:

   probably plenty of people in CSE who can teach or can upskill to teach
   

   at:

   COMP6080 model where I'm LiC and then 3-4 other lecturers
   

   ke:

   "Solving modern programming issues using Rust" as a possible title
   

   zk:

   hard to make it a two-week thing in other courses
   

   tk:

   3141/3161 do a throwaway mention to Rust, but can't do it justice without a separate course
   

   aq:

   frame in terms of concepts, but not PL
   

   jas:

   Are there any objections to moving this forward to the next committee, pending fixes?
   

   Vote Passed




6. New Course: COMP6772 Revisited

   Original proposal for a more advanced C++/SE course was sent back for refinement.

   cjdb:

   key differences to previouss proposalare:
   * documented motivation
   * removed open source component
   * focuses more on C++ in industry
   

   shrey:

   are labs too much?
   

   hs:

   it was really valuable having labs in the first offering of COMP6080, even if they got removed after one term
   

   shrey:

   course weighting inconsistent between secitons
   seems to have addressed issues from last time; good
   

   jas:

   what about the discussion we had last time re integrating this into SWE program?
   

   tfpk:

   proposal has changed quite a bit since last time; you'd need to teach COMP6771 in SWE, which wouldn't make sense
   

   jas:

   okay
   who can take the course?
   

   cjdb:

   anyone, as an elective (no program restrictions)
   

   hs:

   COMP6771 and COMP6772 naming could be improved
   

   shrey:

   ethics being taught in this course is a Good Thing (tm)
   

   jas:

   Are there any objections to moving this forward to the next committee, pending fixes?
   

   Vote Passed




7. Course Offerings 2022

   jas:

   some 1st/2nd year courses have become too big
   suggestion: offered more courses three times a year
   40 courses in T1, 42 in T2, 42 in T3
   120ish courses, 55 academics
   can't cover all the courses, but we have casuals
   

   aq:

   we need to remove courses, can't keep adding
   

   tk:

   don't have enough casuals
   

   ss:

   not equipped for the scale
   

   tk:

   if we don't have enough teachers, it's going to negatively impact UX
   this isn't a problem that's going away
   +400 students/year
   big issue
   in any one term, perhaps courses will be slightly smaller
   school in-fighting over tutor availability
   

   at:

   I don't expect to see much growth in first year
   

   tk:

   large demand on tutors
   

   sri:

   don't think it'll be a problem because we'll always have a top 10%
   

   jj:

   we keep losing tutors to other courses
   

   ss:

   good student does not necessarily mean good tutors
   

   at:

   I don't think there's a problem
   course runs three times, becomes 2/3 the size, so 2/3 the tutors
   

   ss:

   meeting next week should cover this
   

   ab:

   harder to persuade PhD students to teach
   four years of enrollment => faculty wants them to graduate
   

   jas:

   advanced level courses always have difficulty finding tutors
   maybe talk with alumni
   

   at:

   having things online makes things more plausible
   the future will always involve some online courses
   

   hs:

   Dumb question but before 2016 most of the time people's first teaching experience was always tutoring, so what's changed? Do we just have higher standards for casual teaching now?
   

   tk:

   there's a lot of benefit for tutors who are a term out
   a lot of benefit for people to have time to bake as a teacher
   

   zk:

   1511 makes students who want to be tutors first need to be assistant tutors
   

   jj:

   we need more teaching weeks, we don't have enough teaching weeks
   

   at:

   COMP1531 and COMP1521 need to go to three times a year
   

   jas:

   3521 and 3821 offered more
   added an extra offering of ethics (massive demand for an extra offering)
   

   at:

   Ethics we definitely need; algos... do we?
   who to teach?
   

   ww:

   happy to find someone to teach this course (or upskill)
   

   jas:

   Mechatronics now require students to take COMP courses
   




8. Assessment Practices in CSE

   jas:

   some complaints about some exams having more work than can possibly be done in 2h
   need to give a sample exam (just use last year's)
   everyone should get their exam reviewed
   some other schools have an exam review committee that checks for simple things
   - do marks add up?
   - are the questions reasonable?
   - committee needs to be reasonably sized
   - challenge is to test the exams submitted with time to review

   tk:

   Is there an exam prep checklist? I seem to recall somethinglike that?
   

   cm:

   In my previous academic life, that was the norm (the committee). It is a tremendous burden.
   

   tk:

   If not, seems like an excellent idea (rather than a committee, which assumes everyone knows how to check everyone's exams/a big burden)
   

   cm:

   Agreed! We have to make sure we don't "over solve" the problem
   




9. Accreditation

   Accreditation, like Winter, is coming ...




10. Any Other Business

    None

Meeting closed at 11:45am