Consider the following relation: Employees(eid: integer, sal: integer, age: real, did: integer), where attribute eid is the primary key of the relation. There is a hash index on eid and also a B+-tree index on age.

• Total points = 55. Full Mark = 50. • Read each problem statement carefully. Make sure that you answer exactly what is required and that your solution is expressed clearly. • Some notation introduced in class, e.g., the notation we used for true risk, empirical risk, the output of a learning algorithm given a training set, parameter vectors, etc., will be used here directly without being defined again. Please, refer to the notes if you have any confusion about the notation. Also, here k · k will be used to denote the Euclidean (i.e., L2) norm. • You may use any result covered in class, but please cite the result that you use.

Consider the binary image 𝐼𝐼 and the structuring element 𝑆𝑆 below. Do the following by hand (don’t use a computer): a. Label the 4-connected components in 𝐼𝐼. b. Label the 8-connected components in 𝐼𝐼. c. Find the output of morphologically opening 𝐼𝐼 with 𝑆𝑆 – assume the origin of 𝑆𝑆 is at its center. d. Find the output of morphologically closing 𝐼𝐼𝑐𝑐 with 𝑆𝑆̂, where 𝐼𝐼𝑐𝑐 is the background. e. Using the results from (c) and (d) above, show that opening and closing are duals of each other, i.e., (𝐼𝐼 ∘ 𝑆𝑆)𝑐𝑐 = (𝐼𝐼𝑐𝑐 • 𝑆𝑆̂). (an explanation with few words will do).

INSTRUCTIONS TO INVIGILATORS: Please collect all exam question papers and exam answer scripts and retain for school to collect. Candidates must not remove exam question papers.

1. For this question, please create a 2 column by 5 row table, then please put the following code snippets into the respective column. The top row should be labeled “Obfuscated” and “Not Obfuscated” (without the quotes) The color coding indicates a “code pair” one part is obfuscated, and the other part is not obfuscated. Please put them in the table appropriately. (10 points)

• The exam consists of 10 questions for a total of 200 marks • Solutions should be submitted as a pdf, maximum size 20Mb. • Solutions should be submitted via give/webCMS. Multiple submissions are allowed, your last submission before the due date will be the submission marked. • Deadline for submissions is 10am, Tuesday 8 December 2020 (AEDST). • Late submissions will be penalized at 25% (applied to raw mark) per 15 minutes or part thereof. • Existing written materials (i.e. passive resources) may be used, but should be properly referenced. Obtaining, or trying to obtain solutions from active resources (e.g. social media, forums, other people) is a breach of the University’s Plagiarism Policy and will be subject to disciplinary action as outlined in the Student Code of Conduct. • Unless explicitly stated, you should prove/explain/motivate your answers. You are being assessed on your understanding and ability. • Be careful with giving multiple or alternative answers. If you give multiple answers, then we will give you marks only for your worst answer, as this indicates how well you understood the question. • Number of pages in this exam paper: 8, including this cover page.

• Answer questions in the space provided in this examination booklet. The last two pages may be used to continue answers if you run out of space. • Answer ALL questions. See the more detailed grade breakdown on the next page for additional information about what happens if you answer additional questions. • Asymptotic notation can be used to state bounds on running times — but you must use this notation correctly, and as precisely as possible. Computer Science 331 cont’d. ID: Page 2 of 16 Question Score Out Of 1. Operations on Binary Search Trees 10 2. Red-Black Trees 1

The goal for Assignment 3 is to modify your multi-threaded RPC server to implement recursive name resolution, provide immediate persistence for key-value pairs, and improve overall performance and scalability. It must still provide all of the original functionality from earlier assignments, as well as the following additional features. 1. Ability for a variable to contain a variable name or a value. This includes setting and retrieving these names. 2. Recursive name resolution: if a variable contains a name, look up that name and use its value instead. Repeat until you hit a number, or until you exceed a limit. 3. Variables must be persistent before your server acknowledges an operation. We’ll provide details on a simple way to do this; there are faster, but perhaps more complex ways that you may use. 4. Support for millions of variables (key-value pairs). Your hash table will have to support large size for this. 5. Improving performance of your server. Details below. As usual, you must have a design document and writeup along with your README.md in your git repository. Your code must build rpcserver using make.

 Case study is NOT simply a summary of an article/blog post. Instead, it should be a review of the chosen case based on the agenda of our course, i.e. the use of ML in Economics.  Think about your case as a presentation done in front of you, and try to come up with any and all questions you might ask during such presentation. I You will likely not find answers to those questions in the original article/blog post, but you can try answering those questions yourself.  Of course, if your case involves some ML models/techniques that we have not yet covered in our class, make sure to very briefly explain what those are.

The aim of this coursework is to explore the classification of images using convolutional neural networks on a different dataset, CIFAR100 (pronounced as “see far 100”). CIFAR100 consists of 60,000 32 × 32 colour images in 100 classes, with 600 images per class. The first part of the coursework concerns with debugging a “broken” neural network and quantitatively analyzing the problem behind this broken network. The second part involves exploring the solutions in literature for fixing this “broken” neural network, and then subsequently implementing one of these solutions to improve the performance and training of this network. In order to support your experiments, we have acquired Google Cloud Platform credits which allow the use of the Google Compute Engine infrastructure. You will need this to run all the tasks of the coursework (which will be carried out in PyTorch). Each student enrolled on the MLP course will receive a $50 Google Cloud credit coupon which is enough to carry out the experiments required for this coursework. You will receive an email which will give you the URL you will need to access in order to request your Google Cloud Platform coupon. As in the previous coursework, you will need to submit your python code and a report. The detailed submission instructions are given in Section 8.2 – please follow these instructions carefully.