Creative Prediction with Neural Networks

A course in ML/AI for creative expression

How to be a Project Student in the Intelligent Music Lab

Charles Martin - The Australian National University

Ngunnawal & Ngambri & Ngarigu Country

Welcome! Great to have you here!

You’re now part of a research lab (woo hoo!)

It’s a loose network of academics, HDR students, and Master/Undergrad project students!

  • Sometimes we have a formal project (e.g., Charles is my supervisor).
  • Sometimes we have a collaboration (e.g., I’m working on a paper with Yichen and Charles).
  • Sometimes it’s part of a job (e.g., I’m a research assistant working for Henry).
  • Sometimes we just participate (e.g., I’m going to go to Ben’s seminar today to hear about live coding).

What is this lecture about?

  • A brief intro to how work on a project with Charles and others in the lab.

  • Expectations: what we expect of you, what you can expect from us.

  • Practicalities

  • How to plan research

  • Guide to weekly activities

  • Guide to the “shape” of a project.

Problems (for Charles):

  • there are (up to) 10 of you and only one of me!

  • every student needs some of the same information

  • progress on projects is uneven over the semester

  • sometimes students don’t show up

Problems (for you):

  • you have to define a project, execute it, and write it up! (it’s a lot of work - luckily you have 1-2 semesters to do it!)
  • you might be new to research (for now!)
  • you might not have a lot of background knowledge (yet!)
  • research projects are tricky! sometimes ideas don’t work out, you have to be flexible but committed to reaching the goal (handing in a report/artefact/thesis)
  • you’re probably busy! (other subjects, jobs, life!)

The Weekly Status Meeting

How it works:

This is a group meeting, online, once per week. In each meeting you will be asked three questions:

  1. What progress have you made?
  2. What obstacles have you encountered?
  3. What is your next step?

Purpose:

  1. to gain an understanding of how everybody in the lab is going with their project
  2. to determine if anybody requires individual meetings

Status meeting will be only 30 minutes! (approx 3 minutes each!)

Does this look like agile development? Ref: (Hicks and Foster, 2010)

On-demand Individual Meeting (<= 30 minutes)

How it works:

This is an individual meeting, either online or in-person. Book a time through Microsoft Bookings.

Requirements:

Will have a specific goal e.g.:

  • talk about possible solutions to problem X
  • fix technical issue Y
  • discuss an ethics protocol
  • analyse results to an experiment
  • goal should be SMART (Specific, Measurable, Attainable, Realistic, Time-bound)

Student and supervisor should know exactly what we are going to discuss.

Seminars

How it works:

A group meeting for delivering information with relevance to multiple people. This will be dual-delivery! (both in-person and online).

Examples:

  • how to structure a report/thesis in LaTeX
  • how to structure a git repository for project delivery
  • how to apply for an ethics protocol
  • how to run a user study
  • research topics in musical machine learning (mostly from Creative Prediction
  • research topics in sound and music computing (some overlap with COMP2710 Laptop Ensemble)
  • research topics in interactive media (some overlap with COMP1720 Art and Interaction Computing)

Revision: Three types of meeting

  • Status meeting: just for updating status (not research)
  • On-demand meeting: just for discussing research (not status)
  • Seminar: just for 1-to-N information delivery (like this)

The shape of a project

Getting started

  • Defining research topic and objectives
  • Background reading
  • Finding your “unsolved problem”
  • Discovering code frameworks and starting points.

The Work

  • Coding up a system
  • Iterating on a design
  • Running experiments

The Write-up

  • Expressing your process
  • Analysing and discussing results
  • Making images and plots
  • Coming up with your findings and contributions
  • Writing documentation and packaging code

Important Note!

Your grade is based on a combination of your written report, artefact, and presentations.

Examiners look for your ability to:

understand and synthesise existing research

create an artefact, or code project skillfully, and explain it

perform experiments or critically engage with your research problem.

Your grade is not based on your results (really!) (see Honours marking guidelines)

If you get “good” results (yay!) but don’t discuss them in detail, or critically engage in why they are good, you won’t do well. (and the converse)

Project Planning

A bad project plan

  • Getting started: Weeks 1-9
  • The Work: Weeks 10-11
  • The Write-up: Week 12

A better project plan

  • Getting started: Weeks 1-6
  • The Work: Weeks 3-10
  • The Write-up: Week 6 and Weeks 9-12

To help, here’s some deadlines:

1-Semester Project

  • End of week 6: Draft of report Introduction and Background sections (and draft code repository)
  • End of week 8: Draft of report System Design section (and draft code repository)
  • Start of week 11: Draft full report

2-Semester Project.

  • S1: End of week 6: Draft of report Introduction
  • S1: End of week 8: Draft of report Background section
  • S1: End of week 11: Draft code repository and/or report System Design sections
  • S2: End of week 6: Draft of Experimental Methodology and Results sections
  • S2: End of week 8: Full report, draft 1
  • S2: End of week 11: Full report, draft 2

Research Planning