CMPE 257 - Machine Learning

Bernardo Flores

Week 2 - Class 2: ML Fundamentals: Part 1

Motivation

Source: https://xkcd.com/1838/
What is Machine Learning?

- What do you think Machine Learning (ML) is?

- Is ML the same as Artificial Intelligence (AI)?

Artificial Intelligence

“The theory and development of computer systems able to perform tasks that
normally require human intelligence, such as visual perception, speech
recognition, decision-making, and translation between languages.”

- Oxford Dictionary
Machine Learning

“Field of study that gives computers the ability to learn without being explicitly programmed.”

- Arthur Samuel (1959)

Well-posed Learning Problem: “A computer program is said to learn from experience (E)
with respect to same task (T) and some performance measure (P), if its performance on T,
as measured by P, improves with experience E.”
- Tom Mitchell (1998)
ML vs AI

Source:
https://aws.amazon.com/comp
are/the-difference-between-artifi
cial-intelligence-and-machine-lea
rning/#:~:text=The%20goal%20o
f%20any%20AI,analyze%20large
%20volumes%20of%20data.
Machine Learning

- It is very hard to write programs that solve brain processing based

problems (e.g. recognizing a face):
- We don’t exactly know how to program because we don’t exactly know how our brain does
it.

- Instead of writing a program by hand, we collect (lots and lots) of

examples that specify the correct output for a given input.
Machine Learning

- ML takes all the collected examples and produces a program that does the
required task,
- A ML program produced by an learning algorithm may look different form a typical
program,
- But if done right, the program works for new cases as well as the ones we trained it on.
Class Exercise
What is a ‘tree’?
Class Exercise: What is a ‘tree’?

- From the shown image set, tell us which ones are a ‘tree’.

- What was your criteria to choose your selection?

- Go to the following link: pollev.com/bennsaturn

Key Aspects of ML

- There’s a noticeable pattern to be learned about,

- No “easy” way to define/program the pattern to be detected,

- However, there’s lots of data supporting the pattern.

Examples

- Database mining
- Large datasets (e.g. health monitoring records - identifying possible issues)

- Self-customizing programs
- E.g. your customer experience in Amazon, your consumer experience in Netflix,...

Source: D. Ferreira, S. Silva, A. Abelha,

and J. Machado, “Recommendation
System Using Autoencoders,” Applied
Sciences, vol. 10, no. 16, p. 5510, Aug.
2020, doi: 10.3390/app10165510.
ML Challenges

- Lack of Training Data: In order to train an ML-based algorithm, you’ll need hundreds of
examples. However, and in many instances, you simply don’t have access to millions of
real examples representing the exact same thing. That’s the very first challenge machine
learning specialists have to overcome,

- Poor Quality of Data: You may end up having a lot of data, but the quality of it leaves a lot
to be desired. If you start work with poor-quality data, you can’t expect to end up with a
fully functional and effective algorithm. On the contrary, it will be defective and inefficient.
That’s why it is said that the vast majority of the data scientists’ work revolves around
organizing and cleaning data,
Source:
https://addepto.com/blog/what-are-the-to
p-10-challenges-of-machine-learning/
ML Challenges

- Data Overfitting: Data overfitting is all about developing a too complicated machine
learning model and trying to fit it into a limited set of data. In the human world, it’s called
overgeneralization (e.g. food opinion),

- Data Underfitting: The developed model is too simple or misses parameters that it should
have been included in order to produce a clear and unbiased result. This means that your
machine learning model cannot draw useful conclusions from the training data,

Source:
https://addepto.com/blog/what-are-the-to
p-10-challenges-of-machine-learning/
ML Challenges

- Irrelevant Features: Training data is not everything. You also need a good set of features
on which your algorithm can be trained (e.g. spam/not-spam),

- Data Security: You need to make sure that every framework, every third-party app, and
every piece of your IT infrastructure is properly secured against diverse cyber threats,

Source:
https://addepto.com/blog/what-are-the-to
p-10-challenges-of-machine-learning/
ML Challenges

- Cost/Accessibility: When designing a ML solution, how costly will it be? Will it generate
profit yet it is too costly to maintain it? Will you develop open-source? And what about
code practices?,

- Development/Deployment: How will you find the balance between having the right time to
properly develop your ML solution and have it ready for your company to launch? Having a
hasty launch could lead to bugs/issues with the solution but being too ‘slow’ may lead
your customers to another company,

Source:
https://addepto.com/blog/what-are-the-to
p-10-challenges-of-machine-learning/
ML Challenges

Reminder!

The only way to deal with most challenges is to develop an algorithm that’s built strictly
with a specific purpose in mind. Everything has to be adjusted and tailored to your
assignment or project.

Source:
https://addepto.com/blog/what-are-the-to
p-10-challenges-of-machine-learning/
Thank you!