APS360: Applied Fundamentals of Deep Learning

Project Progress Report Handout and Rubric

Introduction
The project progress is a check-in to show that you are on track to complete your project. By the
project progress date, your team should have:
• Obtained all or most of the data
• Come up with a reasonable baseline model
• Produced at least one qualitative and quantitative result
• Provide a convincing argument for the feasibility of completing the project within the time available

How to submit
Submit your project progress report as a group on Quercus by the Progress Report deadline. Any work that
is submitted up to 24 hours past the deadline will receive a 20% grade deduction. No other late work (i.e.
more than 24 hours past the deadline) will be accepted. Quercus submission time will be used, not your
local computer time or any other screenshots that you provide. You can submit your work as many times
as you want before the deadline.

This document should be written using Latex based on the course template. There is a 9-page limit (ICLR
conference standard) for the main text and unlimited pages for references. It should make use of tables,
model diagrams and figures to make the report concise and easy to read. The text should be kept to a
minimum and are specified for each of the three sections that are to be submitted. If any submission exceeds
to 10 pages, a 20% penalty will be applied. We do not accept any submission longer than 10 pages (100%
penalty would be applied).

Progress Report Rubric

The project progress document is graded out of 20 points.

• Brief Project Description (5 points): In your own words provide a brief description of the motivations
behind your project, the goal of your project, why it is interesting or important, and why deep learning
is a reasonable approach.
◦ Someone unfamiliar with the project should be able to quickly establish the project goals, why the
project is interesting and/or useful, and why machine learning is an appropriate tool for the task.
◦ Provide supporting images and diagrams to minimize the required reading.
◦ Visual elements are professional, concise and easy to read.
◦ Make sure you show what is the input and what is the output of your model and justify why deep
learning is a good approach to generate such outputs from your inputs.

• Individual Contributions and Responsibilities (5 points): Provide a summary of how your team
is working together. Describe the project management software that you are using to communicate with
each other, track progress and results, track updates to shared code, etc. You should provide details on
what each person is responsible for and has accomplished so far, along with an updated list of tasks and
deadlines for each member.
◦ It is expected that you will provide a great deal of detail on what the expectations are for each team
member and what they have completed to date.
◦ The progress of the team is well summarized, and it should be easy to determine if the team is on
track, falling behind, or unlikely to complete the project.
◦ The contributions and quality of communication are clearly and concisely documented for each team
member. If a member has not been contributing equally to the team or failing to meet the
deliverables that were agreed to earlier, then this should be made clear in the progress
report.

University of Toronto Faculty of Applied Science and Engineering Page 1 of 2

 APS360: Applied Fundamentals of Deep Learning

◦ An updated project plan and distribution of work is provided.

◦ You are expected to provide redundancies for some special (sensitive) tasks that a member is respon-
sible to do to address the any risks that may prevent you from completing the project, or that plans
for how you can obtain help from your other team members.
◦ The summary is professionally written, concise and easy to read.

• Notable Contribution (10 points) : Provide a detailed summary with results (figures and diagrams)
of the following:

– Data Processing: You can describe the data that you have collected and cleaned. Be clear and
specific when describing what you have done, to the point that someone could reproduce your work.
If possible, show some statistics about your cleaned data (e.g., number of examples in each class), and
at least one example of a cleaned training sample. You should also mention how you plan to obtain
new data for final testing of the model. We prefer you collect some new data yourself or if that is
not possible, use an entirely different dataset which was not used anywhere before.
◦ Clearly describes sources of data, and the steps you took to clean and format your data. Statistics
and data examples are well chosen and show that you have completed data processing. Data
cleaning descriptions are clear enough to be reproduced by a classmate.
◦ Visual elements are professional, concise and easy to read.
◦ Provides a plan for testing on never before seen data.
◦ Describe any challenges you may have faced with the task.

– Baseline Model: You can provide a diagram to describe the baseline model that you have tested and
how it was/will be compared with your primary neural network model. The baseline model can be a
simple machine learning model (e.g., SVM, Random Forests, Models from labs and tutorials, etc.), a
hand-coded heuristic model (that does not use machine learning), or something else. A model that
performs randomly is not acceptable. The expectations for the baseline model will vary from
project to project. You should also provide at least one quantitative and qualitative result. These
results could be learning curves, or results showing the performance of the model on selected samples
of data. The focus here is on assessing the feasibility of your model to achieve the project objectives.
◦ A reasonable choice of baseline for the problem being solved.
◦ The baseline model is easy to implement and requires minimal tuning.
◦ Visual elements are professional, concise and easy to read.
◦ The best results achieved with your model in terms of quantitative (i.e., accuracy, error, loss) and
qualitative (i.e., identify something interesting about how your model performs on select samples
or class of samples).
◦ Describe any challenges you may have faced with the task.

– Primary Model: You can provide a diagram to describe the best model architecture that you have
so far. This description should be more detailed than in your initial proposal. You should provide
a rough idea of how complex your model is (e.g., number of layers, number of parameters, etc.),
and all the details necessary so that someone can reproduce a model like yours that will perform
similarly. You should also provide at least one quantitative and qualitative result. These results
could be learning curves, or results showing the performance of the model on selected samples of
data. The focus here is on assessing the feasibility of your model to achieve the project objectives.
◦ The choice of architecture makes sense for the problem.
◦ The architecture implementation is based on neural networks.
◦ Visual elements are professional, concise and easy to read.
◦ The best results achieved with your model in terms of quantitative (i.e., accuracy, error, loss) and
qualitative (i.e., identify something interesting about how your model performs on select samples
or class of samples).
◦ Describe any challenges you may have faced with the task.

University of Toronto Faculty of Applied Science and Engineering Page 2 of 2