rubric-assignment1
RMIT Classification: Trusted#
Weight Elements HD+ HD DI CR PA PA- NN
A
p
p
ro
ac
h
50%
1) Data exploration leading to well
informed approach.
2) Identifying an adequate evaluation
framework that is tailored to the
problem.
3) Well justified network architecture
and objective.
4) Hyper parameters selection strategy.
5) Approach satisfies all the
requirements and restrictions.
Outstanding across
the course.
The approach is an excellent
and extremely thorough
investigation of the chosen
ML problem. All elements
adequately analysed.
Goes beyond what is done in
class.
Written own data-loader
Single network predict
both outputs
The approach is a good and
reasonably thorough
investigation of the chosen
ML problem.
There are small gaps between
in the investigation in what
could have been explored.
Goes beyond what is done in
class.
Written own data-loader
The approach is sufficient,
but not a thorough
investigation of the chosen
ML problem.
There are gaps in the
investigation and alternative
algorithms or techniques are
better than the ones in the
approach. The approach has a
limited consideration of the
unique aspects of the chosen
ML problem.
The approach is a minimally
sufficient investigation of the
chosen ML problem. It only
examines the bare minimum
requirements of suitable
techniques and algorithms.
There are many gaps in the
investigation and there are
algorithms or techniques are
clearly more suited to the
chosen ML problem.
Poor, superficial,
or incomplete
approach that does
not meet the
minimum
requirements for
PA.
Not Completed
U
lt
im
at
e
Ju
d
gm
en
t
&
A
n
al
ys
is
20%
1) Analysis of the model and the
outputs using suitable methods.
2) Make a clear ultimate Judgment.
3) Rationale behind the ultimate model
is clear and considers all the aspects.
4) Limitations of the model identified.
Outstanding across
the course.
Ultimate Judgement is
established and exceptionally
justified
Evaluation of the Ultimate
Judgement is exceptional and
clearly demonstrated the
viability of the trained model
in real-world practice and
limitations.
Analysed the models output
using independent test data.
Ultimate Judgement is
established and suitably
justified
Evaluation of the Ultimate
Judgement is sound and
suitably explained, however,
the reader may not be fully
convinced and have minor
questions.
Analysed the models output
using independent test data.
Ultimate Judgement is
established, but there are
unexplained choices, or the
justification is hard to
follow.
An sufficient attempt at
evaluating the Ultimate
Judgement is made.
An Ultimate Judgement is
made by not justified.
An Ultimate
Judgement is not
made.
Not Complete
T
es
t
R
es
u
lt
s
20% Performance on the unseen test set
Outstanding across
the course.
(Top 5%)
In Top 5-20% In Top 20-30% In Top 30-50%
Better than random model
but not in top 50%
Less accurate than
random model Not Complete
Im
p
le
m
en
ta
ti
on
&
R
ep
or
t
P
re
se
n
ta
ti
on
10%
1) Code is well documented and easy
to understand.
2) Code does not contain errors.
3) Code contains evidence of all
investigations mentioned in report.
4) Code is optimal and shows good
programming practices.
5) Well structured and easy to read.
6) Reader can fully understand the
rationale for the approach taken
Outstanding across
the course.
Code is exceptional and
satisfy all the elements.
Report is easy to read and
flows well. It is structured
well, leading the reader to
fully understand the rationale
for the final approach taken.
Code is styled and organised
reasonably. Commenting
could be improved.
Report is reasonably easy to
read and flows relatively
well. It is structured
reasonably well, leading the
reader to reasonably
understand the rationale for
the final approach taken.
Code is styled and organised
reasonably. Commenting
could be improved. Few
minor errors.
Report can be followed but
does not flows well in places.
It is adequately structured,
but reader may find it
difficult to understand the
rationale of selected approach.
Code is styled and organised
poorly, not following general
good
programming practices.
Commenting is rare.
Implementation has minor
issues but works.
Report is difficult to follow
and doesn’t flows well.
Readers find it difficult to
understand the rationale of
the selected approach.
Code is styled and
organised
poorly, not
following general
good programming
practices. Contain
major errors.
Incomplete or error
ridden report.
Not Completed