Scribd
Upload
4 views

Final Presentation - AI Stream

Uploaded by

Rich 1st
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
4 views

Final Presentation - AI Stream

Uploaded by

Rich 1st
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 4

School of Engineering Technology and Applied Science

Information and Communication Engineering Technology

Software Development Project II (COMP313)


Final Presentation and Code for 2nd Release (25%)
Due Date: Sunday of Week 13 by 11:59pm (late penalty: 20 pts. per day)

FINAL PRESENTATION REQUIREMENTS & Code


Your team will submit a video presentation of the requirements specified below. Use
slides and live demonstrations of working code. There is no time limit for the video
presentation so please do take your time and do not rush your presentation.

I will be looking for clarity in your presentation when I assign points (i.e., your
understanding of the choices you made in solving the problem you have chosen for
your project). Remember my emphasis on supporting documentation when it
comes to your writing; in your presentation, do not regurgitate what is obvious but
provide rationale for your decision.

1.0 Recap of Midterm Presentation (5 pts.)

Provide a recap of your accomplishments at the end of the midterm presentation (i.e.,
heading into the final presentation).

5 pts. deduction for incorrect, incomplete, nonsensical, or non-submission.

2.0 Hyperparameters and Fine Tuning (20 pts.)

Present and discuss the rationale behind your decisions with respect to the
hyperparameters in your system. Additionally, discuss the methods used to update
weights (or probabilities) in your system so as to maximize your system’s generalization
capability. Please note that the following should be the focus of your discussion.

● The hyperparameters in your system (e.g., number of nodes, number layers,


number of epochs, etc.) [10 pts]
● The approach used to train the dataset to your system to maximize its
generalization capability (i.e., # of epochs, # of iterations, batch, incremental,
online, cross-validation, hybrid methods, etc.) [10 pts]

10 pts. deduction for incorrect, incomplete, nonsensical, or non-submission of each


requirement and associated supporting documentation.
3.0 Evaluator Step: Training and Testing (20 pts.)

Present and discuss the generalization capability of your system. In particular, present
and discuss the following:

● Show visually (i.e., graph) the distribution of your validation and test datasets so
that when you feed your test dataset into your trained system, you can ensure that
the results are not biased [5 pts].
● After you do the above step (i.e., show the distribution of your test dataset), apply
that dataset to your trained system. That is, perform a live code run. If you do not
clearly demonstrate this step, you will get zero for this entire component (i.e., 40
points will be deducted) [5 pts].
● Provide a quantitative analysis (statistical confidence) of your system’s performance
on the test dataset [10 pts].

10 pts. deduction for incorrect, nonsensical, or non-submission of each requirement and


associated supporting documentation.

4.0 Conclusion and Future Works (5 pts.)

Present and discuss the overall results of your project [2.5 pts] and possible future works
[2.5 pts].

5.0 Code (50 pts.)

The following are required with respect to your system’s code:

● All source code files must follow industry standards in terms of format (e.g., proper
indentation, spacing, blank lines, etc). 0.5 pt deduction for every occurrence that
violates these standards. [2.5 pts]
● All source code files contain at the very minimum:
○ at module or class level: original author (citation is very important),
purpose, last modified date
○ at method or function level: purpose, parameters, return value, and side
effects (if applicable)
1 pt deduction for every occurrence that violates these standards. [10 pts]
● Code comment and documentation. Providing comments for code such as loop
counters are not required and in fact, may clutter your code. Use your best
judgment when documenting your code. Remember that code documentation is to
facilitate understanding of code for yourself and others on your team and
organization. 0.5 pt deduction for every occurrence that violates these
requirements. [2.5 pts]
● Data preprocessing code [15 pts].
● Training code. [15 pts]
● Testing code. [5 pts]

You are required to submit your system’s code by giving the course instructor access to
your code repository, preferably GitHub. If you are using something other than GitHub,
make sure to provide the instructor with information on how to access your code. You will
receive a zero if the instructor is not given access before the due date. A zero will also be
given if you submit nonsensical code.

NOTE: relevant supporting documentation in this course refers to text explaining the
rationale behind your decision. For example, let’s say you decide to use One-hot
Encoding as the representative encoding in your project; instead of a descriptive
regurgitation that One-hot Encoding is a vector that allows for only a single ‘1’ and all other
entries contain ‘0’, provide the rationale behind your decision despite the increase in
dimensionality when using One-hot Encoding. This is akin to showing your solution when
solving a mathematical problem instead of providing just a single number or expression as
your answer.
Submitting your work
● Submission is via Ecentennial’s Dropbox.
● Do NOT archive (i.e., zip) your submission. Work submitted as a zipped file will
automatically be deducted by 10 points.
○ Source code should be submitted via code repository by providing access to
the course instructor (see section 5.0).
● For video submissions, a link to the video is required. The video should allow for
streaming. Videos that require me to download it will result in 20 points being
deducted. NOTE: links to videos should not be public links in order to protect the
privacy of all participants.

You might also like