Comparative Intelligence and Ethics in Machine Learning Systems

Jul 11, 2026 15:09 · 29:15 · English · Whisper Turbo · 2 غږوونکي
دا ليکنه نن ختميږي. د تلپاتې ذخيرې لپاره لوړول →
د لوستنې لړيال
يوازې ښودل
0:00
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Hello,
0:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Dr.
0:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Oladako.
0:02
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It's Ben here.
0:03
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I'm here to do a presentation on the
0:07
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
machine learning module.
0:09
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And the topic of the presentation is comparative intelligence and ethics
0:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
in machine learning systems.
0:17
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
This is the agenda of my presentation.
0:20
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I will look at introduction,
0:21
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
purpose of the assignment,
0:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
dataset selection and pre -processing,
0:26
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
selected application track,
0:28
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
model design,
0:29
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
where I look at machine learning and deep learning.
0:33
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
On this topic,
0:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I will look at evaluation matrices,
0:37
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
validation strategy,
0:39
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
explainability findings.
0:41
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And then I'll also go and look at ethical and bias consideration,
0:45
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
deployment consideration,
0:47
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
model demonstration and choice of tools used,
0:50
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
conclusion,
0:51
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
recommendation,
0:52
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then I'll also look at the references used on the presentation.
0:56
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
On the introduction,
0:59
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I just want to say that machine learning
1:03
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
systems have become central to modern decision
1:07
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
making.
1:08
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
From healthcare diagnosis to financial features,
1:11
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
this is supported by REN and others on their paper of
1:16
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
2024.
1:16
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
As these systems evolve towards greater autonomy
1:21
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and complexity,
1:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
questions arise where one wants to check their intelligence compared
1:27
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
to human reasoning and also check their ethical principles.
1:32
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
that needs to be embedded.
1:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
on the on their system design the presentation will look at some of these points
1:38
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
but added on top of this will be other points while some of the points
1:42
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
that i will look at it would be the entire intersection of computational
1:46
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
intelligence and moral responsibility that's what i want to investigate
1:50
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
examine how in a different ml models both classical
1:55
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and deep learning reflect varying degrees of interpretability accountability
1:59
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and fairness and look at algorithm decision making
2:03
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and transparency thereof.
2:05
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
On the
2:09
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
purpose of assignment,
2:10
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I want to indicate that this assignment will critically examine
2:14
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
how different machine learning architectures demonstrate varying
2:18
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
forms of intelligence and ethical alignment.
2:22
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It aims to compare classical algorithms from machine learning
2:26
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and deep learning models in terms of the following.
2:29
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
So we will look at these dimensions.
2:32
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the transparency,
2:33
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
efficiency,
2:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and explainability,
2:36
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
fairness, and accountability,
2:37
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
where this will be looked against the classical ML
2:42
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
models and the deep learning.
2:44
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And these are the characteristics of how these models behave when
2:49
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
you compare them across these four dimensions.
2:53
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Data set selection and pre -processing.
2:57
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The data set that I've selected is the tuberculosis,
3:00
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the TB data set.
3:02
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The reasons why this was chosen is because TB remains
3:06
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
as a major global health issue,
3:08
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and the chest X -rays are a primary diagnostic tool
3:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
to diagnose TB,
3:14
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
use X -rays,
3:16
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and using the data set that was downloaded from cackle .com.
3:20
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
This TB dataset allows ML and DL models
3:25
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
to tackle a real -world medical challenge.
3:29
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And the reason why this dataset was selected is as follows.
3:33
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The dataset itself,
3:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it has a clinical relevance.
3:38
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Why?
3:39
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Because TB is a global health issue and a challenge.
3:42
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
So doing this dataset will help a lot.
3:45
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It has X -rays.
3:46
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It already has X -rays.
3:50
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
diagnosis it is it has a large scale availability it has
3:54
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
a lot of of x -rays embedded on the on the data set itself
3:58
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
thousands and thousands of them it also support deep learning models
4:03
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
so it is very key for this presentation it also has a balanced representation
4:08
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
why balance presentation a representation because it has both the
4:12
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
positive cases and the negative cases for tb it also supports
4:16
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
a supervised models and then
4:19
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It has a good potential on benchmarking
4:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
because it is widely cited in many research papers,
4:27
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
so it is very relevant for this presentation.
4:30
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It also supports the comparison of computer vision methods.
4:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And lastly,
4:36
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it has a strong public accessibility.
4:38
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It is an open source data set,
4:41
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
so it is very accessible,
4:42
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and it also promotes reproducibility.
4:47
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and collaboration as it is an open source dataset.
4:50
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Dataset selection and preprocessing.
4:55
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Dataset preprocessing was done
5:00
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
on the data set to carry
5:04
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
out and prepare the structure activities before the data
5:09
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
analysis on both the traditional LM and
5:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the deep learning.
5:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
On the traditional LM,
5:15
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
these are the pre -processing activities that were applied,
5:19
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
grayscaling,
5:20
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
noise reduction,
5:21
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
histogram equalization,
5:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
dimensional reduction.
5:25
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then on the deep learning resizing was done to standardize the input
5:29
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
size normalization to scale down the pixels values for
5:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
stable for stable training and data augmentation
5:39
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The next slide talks to the selection of the application track.
5:43
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
There were two tracks that were given to us to select from,
5:46
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and we were given the computer vision and the recommender
5:51
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
systems.
5:51
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I just want to indicate that before one chooses a
5:56
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
tracker,
5:57
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
One need to make sure that the tracker that
6:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
you choose will depend on the type of problems that one wants
6:05
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
to solve and the impact that that tracker
6:09
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
will have on the industry that you aim to analyze.
6:14
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
So I did a comparison between the two trackers,
6:17
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the computer vision and the recommender systems,
6:20
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then the focus on the...
6:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
CVS,
6:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the computer vision.
6:25
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I checked the focus.
6:27
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
This tracker always focuses
6:31
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
on image and video analysis.
6:33
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Where is it applicable?
6:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It is always applicable on self -driven cars,
6:37
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
medical imaging,
6:39
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and industrial automation.
6:40
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The workflow,
6:42
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
you get the images as an input.
6:46
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
You do object detection,
6:48
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
you do analysis,
6:49
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then you provide
6:51
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and the understanding of the output as the results.
6:55
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The strength of the CV model,
6:57
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it is very strong.
7:00
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
because it is used well for interaction,
7:03
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
like I indicated,
7:05
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
on self -driven cars.
7:07
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And it also has a very strong support for robotics
7:11
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and safety.
7:12
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The recommender systems,
7:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
where is their focus?
7:15
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Their focus is on personalized recommendation.
7:18
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Where is it applicable?
7:20
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It is mostly applicable on e -commerce,
7:22
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
streaming services like Netflix.
7:25
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
social network like WhatsApp.
7:29
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And the workflow,
7:30
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
you get user data,
7:32
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
then you process that data,
7:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
then based on the data,
7:35
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
you choose the preferred model,
7:38
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
then you analyze the data,
7:40
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then you provide the recommendation as outputs.
7:43
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And then the strength,
7:44
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
where is it strong?
7:45
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It is very strong on user engagement,
7:47
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
like on social media.
7:49
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It is also very strong on sales and retention.
7:52
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Based on the comparison above,
7:54
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
computer vision was chosen for this presentation because the
7:58
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
data set that is used on this presentation is medical images,
8:03
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then it is discovered that CV models on medical images
8:07
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
is one of the most transformative applications of AI
8:11
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
that bridges healthcare and technology to improve diagnosis,
8:16
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
treatment,
8:16
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and patient outcome.
8:18
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
This is supported by Esteva.
8:20
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and others on their paper of 2021.
8:22
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Model design.
8:25
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
On the model design,
8:26
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I looked at the valuation matrices on
8:31
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the traditional ML.
8:35
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
machine learning model,
8:36
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I chose the support vector machines,
8:40
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the SVM model with handcrafted features.
8:43
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
On the deep learning model,
8:44
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I chose the convolutional and neural network,
8:48
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the CNN.
8:49
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Now looking at the results,
8:51
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
classification matrix is one of the results using the SVM
8:55
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
on ML.
8:56
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It has these classifications.
8:58
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the cases where there's no TB,
9:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
zero cases,
9:02
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
cases where there's TB and accuracy and macro average and weighted
9:06
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
average.
9:06
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Looking at the cases with TBU,
9:09
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it used 706 features.
9:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
On the cases with TB,
9:16
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it used 134.
9:18
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And then one conclude,
9:20
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
looking at these results,
9:22
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
one concluded that the SVM model is excellent in predicting
9:26
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
cases where there's no TB.
9:28
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
but reasonably good in predicting cases where there's TB.
9:32
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The overall accuracy is around 95%,
9:36
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
but this shows a slightly lower scores on the classes
9:40
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
where there's TB.
9:41
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And then this suggests that the model struggles a bit with the minority detection.
9:46
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
But this is common.
9:48
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
This is a common issue in data sets that
9:52
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
have imbalances,
9:52
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
especially with the X -ray data set.
9:57
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
continue with the results now we look at
10:00
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
the CNN model,
10:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then it used 10 epochs to
10:05
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
train the model.
10:07
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
And when you look at the accuracy,
10:08
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the accuracy is 100 % across the 10 epochs,
10:12
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then the loss is also zero across the epochs.
10:16
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
What does that mean?
10:17
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
That indicates that the model has no error on training
10:21
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the data,
10:23
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
on training the model,
10:24
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and it also predicts perfectly.
10:27
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
with the match labels now comparing both now the the models
10:32
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the svm from ml models the cnn from
10:36
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
the deep learning
10:37
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Compare the results across these aspects,
10:41
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the accuracy,
10:42
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the loss,
10:43
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the class balance,
10:43
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then see how the tool performed.
10:46
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
On the accuracy,
10:47
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the overall performance of the SVM is around 95%.
10:51
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And then on the CNN is
10:55
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
around 100 % training.
10:57
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And then the loss is zero for CNN,
11:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
which is a perfect fit.
11:03
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
svm is not strong but definitely it is greater than zero on class
11:08
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
balance it is different discovered that it is very strong on the majority
11:12
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
where there are no no tb cases but very weak but a bit
11:16
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
weak on the on the minority where there's tb to classify those and
11:21
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
to predict those and then the on the on the same end there's no distinction
11:25
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it looks like everything is perfect on the generalization the
11:29
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
generalization of the results it is moderate
11:32
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
to with realistic errors,
11:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
but on the CNN,
11:36
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it looks like it has a high risk of overfitting
11:40
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
the model.
11:41
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
On the interpretation of the results,
11:43
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it is really reliable,
11:45
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
but imperfect.
11:46
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The CNN,
11:48
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it is very suspicious.
11:50
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
One needs to further conduct further analysis,
11:53
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
like validation or check metrics.
11:57
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
Continuing with the model design,
11:59
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I'm looking at the validation strategy.
12:02
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I compared both the SVM model
12:06
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and the CNN model using the K -fold method.
12:09
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Then look at the accuracy and the accuracy
12:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
of the CNN.
12:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The CNN was trained on four training folds,
12:18
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and the accuracy of those five folds
12:22
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
training folds,
12:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it is between 95 % and 98%.
12:27
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
of the five folds,
12:28
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
training folds,
12:29
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then the average is 97%.
12:32
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Comparing the two,
12:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it shows that the CNN model performed better
12:38
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
than the SVM because it is 97 % while
12:42
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the SVM is sitting at 94%.
12:45
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
Now,
12:46
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
what that says is...
12:48
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the medical imaging tasks such as TB chest x
12:53
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
-rays classification where data sets are often limited and in balance.
12:57
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Robust validation is critical for building confidence that
13:02
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
both SVM and CN models will perform consistently
13:06
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
in real -world scenarios.
13:08
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
So it is very important to perform this validation using the k
13:12
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
-fold method to ensure that both this model will perform.
13:16
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
accurate on the real -world scenarios.
13:18
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The validation in this case is very important based
13:22
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
on the following reasons.
13:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It ensures that the model performs performance
13:28
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
reflects the ability to generalize to unseen data rather
13:32
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
than just the memorization during the training set.
13:35
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It also makes sure that using strategies like K -Fault
13:40
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
every data point is tested across different splits giving
13:45
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
a more reliable estimate and accuracy.
13:49
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
This process helps in detecting overfit like
13:53
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
in the CNN model where model just performs very well
13:57
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
on training data but fails when it comes to the real sample.
14:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
In summary on this method
14:03
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
One can say that validation acts as a safeguard that distinguishes
14:08
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
genuine learning from memorization,
14:10
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
ensuring that models perform reliably in real -world
14:14
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
scenarios rather than just excelling on training datasets.
14:19
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Now, going to the explainability findings,
14:22
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I need to introduce this topic just to say that explainability
14:27
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
in the machine learning and deep learning models,
14:30
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
especially when one uses the SVM and the CNN model,
14:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
is very essential for transparency,
14:36
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
trust, and ethical deployment.
14:38
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
So I applied sharp on the SVM
14:43
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
model.
14:44
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
to check explainability.
14:46
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
So the sharp summary plot on the right here highlights how
14:50
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
different features contributed to the model's prediction.
14:54
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
So if you look at this one,
14:56
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it shows that a feature like 136.
15:00
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
337,
15:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
505,
15:02
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and 3094 have
15:07
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
negative sharp values,
15:09
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
meaning that they generally reduce the predicted output rather
15:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
than increasing it.
15:15
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Overall,
15:15
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
on this visualization,
15:17
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
One can say that the SVM model is heavily driven by few key
15:21
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
features with their values consistently pulling the prediction
15:26
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
downwards instead of pushing it downwards.
15:29
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
One then can conclude that it is critical for one to understand.
15:33
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
This plot helps
15:39
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
us to understand the model behavior,
15:41
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the diagnosis biases.
15:43
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and to ensure that transparency is there during
15:48
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
deployment.
15:48
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Now looking at explainability for the CNN
15:53
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
model, I applied the GRANDCAM
15:57
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
method and these are the results that came from the
16:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
GRANDCAM method when
16:05
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
applied on the CNN.
16:07
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And this is applied on the chest X -rays,
16:11
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
which visually explains how the CNN model makes prediction.
16:15
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
So when one looks at this,
16:17
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
you see there are two different colors.
16:19
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The dark blue colors,
16:20
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
these are the areas where there's high activation.
16:24
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And this shows that this is where the model forecast
16:29
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
when it was doing its predictions.
16:31
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
on the other hand you see these yellow colors here and these
16:35
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
yellow colors are the areas where there's low activation
16:39
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and these are the cases where there's little contribution to
16:44
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the cnn output
16:46
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
One can then conclude that this visualization is very important because
16:50
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it provides us with the explainability of the model showing
16:54
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
clinicians and researchers where the model forecast when
16:59
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it was doing its prediction for judgment.
17:04
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The next slide,
17:05
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
look at ethical and bias consideration.
17:07
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Bias consideration,
17:10
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
these are the items that I looked at.
17:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I looked at hidden disparities for overall accuracy
17:17
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
to make sure that the subgroups are masked correctly
17:21
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
to cater for differences.
17:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Computer vision on bias,
17:26
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it's very important,
17:27
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
especially for CNN because CNN might perform very well.
17:31
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
on on on on light images but struggles on
17:35
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the on the low contrast images and then recommend
17:40
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
recommender systems these models will will perform very
17:45
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
well on high activity users compared to low user activities
17:50
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
so one needs to take that care when you do the the the bias
17:54
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
one subgroup analysis
17:56
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The results need to be broken down into demographic,
17:59
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
environment and behaviour factors to make sure that the generalisation
18:04
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
are clear where the metal performed very well to generalise
18:08
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and where it failed.
18:09
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Critical role needs to be defined
18:14
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
on the model to make sure that fairness,
18:16
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
trust and safety is there to safeguard
18:21
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the reinforcement.
18:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
of inequalities and to ensure that compliance with
18:27
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
ethical standards is achieved.
18:30
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Improvement strategies,
18:31
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
always make sure that improvement strategies are there on the models.
18:35
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
One can always use techniques like
18:40
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
data augmentation,
18:41
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
reweighting,
18:42
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
fairness, awareness training,
18:44
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
then to help to mitigate against the disparities.
18:49
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Ethical consideration continues,
18:51
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
but now I'm looking at ethical considerations.
18:55
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And then these are the issues that I looked at when
18:59
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I looked at ethical consideration.
19:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
I looked at data privacy to ensure that originally the
19:05
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
concern and anonymization of sensitive data or personal information
19:09
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
is adhered to algorithm fairness.
19:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
This ensures that balance false positive and
19:18
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
false negative are avoided so that the harmful issues
19:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
are taken care of in the real
19:27
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
world application of the models.
19:30
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Bias and harm,
19:31
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
this needs to be addressed because it will mitigate
19:36
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
against the risk for bias data and model decision.
19:40
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
when you apply this to cater for inequalities.
19:45
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Ethical reflection always makes sure that when
19:50
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the model is applied there's always
19:54
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
that ethical reflection to ensure that fairness in
19:58
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
specific use cases
20:00
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and recommend models are always there to validate across
20:04
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
subgroups for equal deployment.
20:06
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The next slide looks at the
20:11
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
model deployment.
20:12
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
We are not required to deploy the model per se,
20:15
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
but the requirement was that
20:18
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
we provide the consideration how the model will be deployed.
20:21
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
This is the workflow that I've developed to deploy the model.
20:25
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The first step is to do data collection and processing.
20:29
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
On the data collection and processing,
20:31
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
one gets data from different sources.
20:34
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
You do data cleaning using ETA strategies
20:38
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
or methods,
20:39
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then one also apply feature engineering on the data set.
20:43
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Once this is done,
20:44
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the clean data is then passed to the model training and validation.
20:48
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The model,
20:49
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the validation and training,
20:51
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
this is where the model is trained on the data.
20:54
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
You do validation of the model where you use different matrices,
20:59
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
explainability or performance.
21:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And then you also check for fairness checks.
21:03
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Once this is done,
21:05
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
then you pass the model to the deployment.
21:07
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
This is where then the model is deployed.
21:09
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And it is recommended that you use it when the cloud platform
21:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
is used, for example,
21:15
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
like Azure.
21:15
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
When deploying the model,
21:18
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
one can also deploy it with APIs.
21:22
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Once the model is deployed,
21:23
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
then monitoring and scaling is applied,
21:26
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then this is where you track your matrices to auto
21:30
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
-scaling to improve performance,
21:32
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and then also check bias and drifting alerts on the model.
21:36
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Once that is done,
21:37
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
then based on the output or the results from the monitoring
21:41
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
and scaling,
21:42
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
One is to retrain the model.
21:44
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
How do you then retrain the model?
21:47
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
You go back and connect more data to feed into the
21:51
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
model, then based on that you update the model and then you
21:55
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
go back,
21:56
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
retrain the model.
21:57
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
That's how I propose the deployment of the model.
22:01
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The next slide looks at the piece of
22:06
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the notebook that I've developed,
22:09
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
just to indicate that the model does perform,
22:12
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
does work.
22:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
This is just a demonstration on how the model works.
22:16
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
Yes,
22:40
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
this is how the model works.
22:42
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
So this is how I developed the model.
22:47
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And this piece of the model,
22:48
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
actually what it does is it is
22:53
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
from the notebook,
22:53
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
the entire notebook.
22:55
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Then how does it perform?
22:56
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It does the data pre -preparation and then it does
23:00
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the visualization pipeline for the chest X -rays.
23:04
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
What does it do?
23:05
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
How does it do that?
23:06
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It loads the images.
23:08
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
pre -process the images,
23:10
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
create labels,
23:12
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
arrays,
23:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
applies the data augmentation,
23:15
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
and finally produces the visualization of the chest
23:19
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
X -rays.
23:20
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
That should show that the entire model does work and it does produce
23:24
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the visualization of the X -rays.
24:13
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The next slide looks at the choice of the tools.
24:17
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Why did I choose the tools that I chose for the model?
24:20
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The SCETEC Learn,
24:23
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it was applied on classical
24:27
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
models.
24:28
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
The reason why it was applied,
24:31
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
it was not because it is an advanced tool,
24:33
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
but it was applied because it provides that transparency to
24:38
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
compare the deep learning pipeline against the...
24:41
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
the classical machine learning model.
24:44
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
So it provides that comparison between the two.
24:48
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Why TensorFlow on deep learning?
24:51
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
TensorFlow on deep learning was used to show how the DL models
24:55
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
can learn features automatically and scale the
24:59
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
large dataset with extensions and transfer learning.
25:03
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
Why use SHARP unexplainability?
25:06
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
SHARP was used on explainability for the following reasons.
25:10
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
It makes black box.
25:12
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
That means it allows the comparison of the
25:16
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
SVM and the CNN results to be interpretable.
25:20
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And then it helps with understanding why the prediction was
25:25
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
made, not just to know what the prediction is.
25:28
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
So it gives you that understanding why was this prediction made.
25:31
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
And then it also provides mathematical consistency
25:35
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
features,
25:36
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
attributes.
25:37
S… Speaker 1 (Comparative Intelligence and Ethics in Machine Learning Systems)
based on the Shapley values.
25:38
S… Speaker 2 (Comparative Intelligence and Ethics in Machine Learning Systems)
And lastly,

دا نقل د AI لخوا رامینځته شوی و (د خبرو اتومات پیژندنه). ممکن غلطۍ ولري - د مهم کارولو لپاره د اصلي غږ سره سم تایید کړئ. AI پالېسه

❤️ STT.ai مینه؟ خپل ملګرو ته ووایاست!
لنډيز
د دې د نقل د AI لنډیز توليد خلاصه کېکاږئ.
...لنډيز کيږي
د دې د نقل په اړه AI وپوښتئ
د دې نقل په اړه هرڅه وپوښتئ - AI به اړونده برخې او ځواب ومومي.
بل لېښل