I've made some minor updates to ass3.pl. Please note that it's not possible to write a comprehensive English grammar with just a few rules. So there are many sentences that can't be parsed. For example, you can write "John gave the book to Mary", but not "John gave Mary the book".
If the grammar can't parse sentence properly, it won't be tested. You can come up with your own test examples by trying the grammar and seeing what it can do.
You will have to have submitted a credible attempt at all three assignments and have a satisfactory mark for at least two. If the third is not considered to be credible, that will not be satisfactory overall. By credible, I mean that we don’t want you to just submit anything just to register a submission. We will have to judge it to be a real attempt.
Assignment 1 marks should be out and assignment 2, I hope will be out at the end of this week.
I'll hold another consultation session for assignment 3 tomorrow (7 May) at 3pm on Blackboard Collaborate.
I will be on Blackboard Collaborate today at 2:30pm for an hour to answer questions about the assignment. Use the Help Session link under the Collaborate link.
I've posted assignment 3. Don't panic when you see it. I've provided a fairly complicated DCG grammar, but you only have to write an additional Prolog program to take a parsed sentence and work out pronoun references.
I will record an explanation of the code and next week, and I'll have an online consultation session to answer your questions. Time to be announced, but mostly likely on Thursday. That should give you time to make a start.
I’ve posted the PDF of the almost last lecture. This one is on grammars in Prolog and you will be using this in the final assignment. I will be recording the voice over this evening. This is the last formal lecture. On Friday, I will post another talk about the final assignment.
Update: The lecture recording is now online.
Some people are reporting that the are getting very bug decision trees. This is probably because they haven't turned on the pruning parameters. So don't forget to prune the trees.
The PDF and lecture recording the second Natural Language Processing (NLP) lecture one Semantics and Pragmatics is online under "Week 9" lectures.
If you are struggling with sklearn in assignment 2, there is an alternative ML toolkit, Weka , that you can use. Either will be acceptable for the assignment submission. If you are not familiar with Python, Weka may be easier because everything is done via a GUI. Weka's J48 algorithm which is a re-implementation of C4.5 also handles discrete variables more cleanly.
You can find a brief explanation of how to use Weka under "Assignments" and linked here .
This week we are starting on Natural Language Processing. The first lecture is on grammars and parsing. The PDF of the is up now and I'll do the recording tomorrow (Tuesday), subject to outside noise.
Update:
The lecture recording is now up and I fixed the permissions so you can see both the PDF and recording.
I put up a new lecture on Inductive Logic Programming over the weekend and the week 9 tutorial on neural networks and ILP is also up now.
We will be moving on the natural language processing next.
Friday is a public holiday. If you have a tut on Friday, you may try to join a Thursday tut instead.
I will still be putting up the final machine learning lecture sometime on Friday.
I've uploaded the slides for the week 8 lecture on neural nets. I will be recording the voiceover today, provided the construction noise outside my place is not too load!
Update:
The video of the lecture is now up.
UNSW has decided that courses will only have a Satisfactory/Unsatisfactory outcome. I know this will disappoint many of you but we have to follow university policy. It's not practical to have an exam, so I will add an extra assignment to be done during what would have been the exam period. None of this is ideal. We are trying to do the best we can under usual circumstances.
There are two options for the final assessment and I'd like to hear your opinions. I can weight the assignments 20+30+50, or make them all equal weight. Which do you prefer? I'd like to hear from you first, and perhaps run a poll later.
Update:
It looks like equal weight has the majority.
I've uploaded the PDF of the slides for the next lecture on decision trees and there's also a link to the video for the lecture under the Week 7 panel when you click on the lectures link in the left sidebar.
The PDF of the Tuesday lecture is now up and the video Is currently uploading and will be available later today.
Note a small difference between the PDF and the video on slide 25 (Value Function). In the video, I forgot to mention that the Q learning algorithm makes a simplification the actions are deterministic, whereas easier I talked about the more general case when they are not.
These videos will be shorted than a normal lecture, because there are not pauses, as I would normally do in a live lecture, so please use your pause button to take in what's on the slide and go back as often as you need to to re-listen to a slide.
More videos on Machine Learning will go up later in the week. The next one will start us on Decision Trees, which are in the second assignment.
I've posted a draft of assignment 2. This is a bit shorter to allow time for a third assignment that you will do during what would have been the exam period. Please not the unusual due date: Tuesday 21 April .
Remember that we are resuming lectures and tutorials this week, all online.
Just a reminder that next week, we'll go back to a more or less normal schedule. I'll release lecture recordings on Tuesdays and Fridays and your tutorials will be at their usual times but please go to Moodle/Collaborate and join your group.
Please make an effort to keep attending regularly and participate actively. That's the only way your studies will still have value in the present circumstances.
Because of the disruption due to the COVID-19 outbreak, the university is considering how to handle the final assessment for Term 1. One possibility is to go to a simple Satisfactory/Unsatisfactory mark. I appreciate that this doesn't reflect the work that you are putting into your studies, but we can't run a normal exam, and it's impossible for us to invigilate any assessment.
I need to consider whether to go for a 24 hour "take home" exam or add another assignment. This may also mean reweighting the marks for the assignments. If you have any comments or suggestions, please respond below, or if you want to comment privately, email cs3411@cse.unsw.edu.au. A decision will have to be made fairly quickly.
Thanks for you understanding. This is a difficult situation for everyone.
For the rest of this term, we are going to run all tutorials online. Please read the information available on the following page to actively participate in a remote tutorial: Online Tutorials
Your online tutorial is at the same time and day as your face-to-face tutorial in the Timetable .
There is no tutorial this week (week 6), however, your tutors will still be on Collaborate - Online Teaching at your usual tutorial times to help you with any queries/questions you might have. It's also a great opportunity to be familiar with this new online teaching platform.
Please do not come to the lecture theatre on Friday. I will be uploading a video of the slides with a voice over. This will not be on Echo360. I will post a link when it's ready.
Update:
The lecture video is now online. Please follow the link under the "Week 5 Lectures" panel. Don't expect a professional production, they have to done pretty quickly, so the videos will be the slideshow with me talking over them with explanations. I hope this will work.
Today help session will be run online via Blackboard Collaborate on Moodle. Click "Help Sessions" link on the left panel or here for more detail.
Assignment 1 spec has been updated with the submission instruction. Your code will be run on a few tests when you submit. So, it is a good idea to submit early and often so that potential problems with your code can be detected early. You can submit as many times as you like - later submissions will overwrite earlier ones.
As a precaution against possible infection, UNSW is asking us to move courses online as much as possible. Today's lecture will go ahead, as normal, but you may wish to view it online instead of attending. We are working out how to take tutorials online, probably using Blackboard Collaborate, which can be accessed via Moodle. More instructions will follow.
For today we'll continue as normal, but please stand-by for more announcements.
Hi everyone,
Information about help sessions can be found here . First help session is today from 5pm-6pm. Times and locations are subject to change, so please check the page frequently just before you head to one.
I've posted additional notes on variable and value selection for the Australian states example, under week 3 lectures.
I've just put up a draft of assignment 1. Please look under the Assignments menu on the left.
Please note: Give has not yet been set up for submissions. I will send a new notice when that's been done and update the assignment spec so you can see the submission instructions.
I've uploaded a new PDF of the slides on uninformed search with corrections to the diagram on slide 12 (breadth-first search) and there was a typo on slide 22 for the space complexity of depth-first search.
In the lecture I said that the lowest-cost search is more efficient than the bread-first search. This was a mistake. The purpose of the lowest-cost search (also called uniform cost search) is to find the optimal solution when the path has a cost, i.e. we don't care about path length, only the total path cost. That is why we replace the depth term, d, in the complexity estimate, by the ratio of the cost of the optimal path, C*, and \epsilon, which is the least cost for one step. In fact, the lowest-cost search can have much worse complexity than breadth-first, but we are trying the lowest cost rather than the shortest path.
The website for SWI Prolog is https://www.swi-prolog.org/download/stable . Please read the installation notes carefully.
If you want to try the SWISH website go to https://swish.swi-prolog.org .
The recordings for today's lecture and tutorial 1 is now available. You can access the recordings:
The lecture recordings are made automatically by the UNSW IT service. They should be available sometime after each lecture.
One tutorial will be recorded a week and uploaded to Moodle until the travel ban ends.
The problem with linking to Echo360 should be fixed today. The lectures should be recorded, even if the link is not in place immediately.
Update:
The lecture recordings link should be working now. Recordings of some of the tutorials willl go up later in the week.
We are currently having some problems linking the Echo360 video recordings. Hopefully this will be resolved by tomorrow, but in case the problem persists, I've put all this week's lecture slides up. Please check the Lectures button in the panel on the left.
Welcome to COMP3411/9814 - Artificial Intelligence.
Please go to the course outline (in the panel on the left) to see the locations of the tutorials and who your tutors are.
If you have any general questions, please post them here. If you are making a more specific enquiry, please email the class account <cs3411@cse.unsw.edu.au>, not my personal email. That way, if I'm not able to answer promptly, someone else can pick up your query.