0% found this document useful (0 votes)
10 views

IntroToMachineLearning - 25-07-2019

Uploaded by

Aliaa Tarek Ali
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
10 views

IntroToMachineLearning - 25-07-2019

Uploaded by

Aliaa Tarek Ali
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 37

Intro.

to Machine Learning
By Eng.Yousra Hesham
Table of Content
• What’s Machine Learning?
• Machine Learning VS Traditional Programming.
• Machine Learning is the future.
• Machine Learning Algorithms.
• Machine Learning Applications.
• How to start in ML?
• Machine Learning process.
• Real Life examples.
• Questions?
• Refernces
What’s Machine Learning?
• How to make Machines Learn  ?
• Simulates the way human Learn.
• Rule-Based Learning.
• Previous Experience-based learning.
• Example: Teach a kid how to identify Cats?

What’s Machine Learning?
• Is it Artificial Intelligence?
What’s Machine Learning?
• Instead of telling the Machine exactly what to do, we feed the data to
a generic algorithm and it builds its own logic.

• Example: Spam checker in e-mails.


Machine Learning VS Traditional Programming

• Traditional Programming: • Machine Learning:


Teach a kid how to deal Teach a kid how to identify
with hot ovens? animals?
If the oven temperature is Based on some features?!!
up don’t get near to it. Or perhaps, by showing a
set of animals so
unconsciously he learns.
Machine Learning is the Future
• Data is every where.

• In 2005 the data was 130


Exabyte and counting.

• Data is ( Songs, Movies,


Books, Articles, spoken
words, text messages…
etc)
Machine Learning Algorithms

• Divided based on type of input Data:


• Supervised Learning Algorithms.
• Unsupervised Learning Algorithms.
• Semi-Supervised Learning Algorithm.
• Reinforcement Learning Algorithms.
Machine Learning Algorithms
• Supervised Learning:
• We have data and know the answer.
• Types:
• Regression.
• Classification.
Machine Learning Algorithms

• Unsupervised Learning:
• We have data and don’t know the
answer.
• Example: distinguish between cats and dogs.
Machine Learning Algorithms
Machine Learning Algorithms
• Semi-Supervised learning:
• The data is divided into 2 parts:
• Few is labeled data.
• Lots of unlabeled data.
• Mix between Supervised and Unsupervised learning.
Machine Learning Algorithms
• Reinforcement Learning:
• Train itself from the environment.
• Usually used with robotics applications.
Work out your mind!

What is the output of Machine Learning


process?
Machine Learning Applications
• Natural Language Processing:
• QA.
• Sentiment Analysis.
• Crowd Monitoring.
• Machine Translation.
• public opinion mining.
• Computer Vision Applications:
• Face Recognition.
• Object Recognition.
• OCR.
• Bio-informatics applications.
• Hand Writing Recognition.
• Facial Recognition.

• Recommender Systems ((Amazon- Facebook - Netflix-… etc).


• Virtual Reality headsets.
How to start in Machine Learning?

Build a basic foundation + Practical knowledge in


a chosen track
How to start in Machine Learning?
• Build a basic foundation:
• Learn the basic theory (Maths) behind ML.
• Python or R.
• Understand ML Algorithms.
• Practical applications along the way.
• Join competitions ( why not? )
• Resources:
• Kaggle Crash courses. (amazing for a strong hands-on start)
• Machine Learning A to Z by Data Science (for overview and practice)
• Intro. To ML by Andrew NG (for theory)
• Machine Learning Nanodegree by Udacity.
Steps towards an ML Model
Steps towards an ML Model
Lets take Examples, shall we?
Natural Language Processing (NLP)
• Sub-field of Artificial Intelligence.
• Enabling computers to understand and process human languages.
• Example:
• Apple Siri.
• Amazon Alexa.
• Google Translate.
• Chat Bots.
Classify Restaurant Reviews
• We have people’s reviews on a restaurant and we want to predict if others are
Positive or Negative.
• Wow….. I loved this Place. -> Positive

• Procedure?
• Remove the non-letters characters. Wow i loved this place.
• Lower Case letters. wow….. i loved this place.
• Make the text sentence to array of words. [ wow, I, love, this, place]
• Removes the unnecessary parts (Pronoun–Stop words....( . [wow, loved, place]
• Get the origin of a word(Stemming). Loved -> love. [wow, love, place]
• Choose a suitable model for the problem!
• Test your model with new data( other than the training Data)
• Calculate it’s accuracy, if acceptable then you have your model ready, if not , re-ajust the
model parameters of perhaps a new model.
Machine Translation
• Making machines translate from one language to another.
• This field is part of NLP.
• Two approaches:
• Old approach: ‫أنا أحب كرة اليد‬
• Depends on the linguistic rules of the Language.
• New approach ( Statistical):
• Depends on the already existing data.
I love Volley ball
Statistical Machine Translation

• Steps:
• Break original sentence
into chunks.
• Find all possible translations for
each chunk.
‫أنا أحب الذهاب الي أجمل شاطئ‬
• Generate all possible sentences
and find the most likely one.
• Limitations?
No available dataset from each
languages to another so, we use
an in-between language.
Statistical Machine Translation

• Make it even smarter:


• Recurrent Neural Networks
• Use the previous word in the
sentence to predict the next one
translation.
Statistical Machine Translation

• Make it even smarter:


• Encodings
• Turning raw data into the list of
measurements that represent it.
Machine Learning without Machine Learning
Knowledge
• Can someone develop an ML program and let me use it?
• Astonishingly, the answer is YES. -> Google APIs is an Example
• Vision API, NLP API… etc.
• You now has a tool (applicable with no fee for small startups) to start focusing
on your application idea not the implementation of the ML Model.
That’s it for Today?
Questions?
Arabic Tokenizer
• the process of breaking a stream of text up into words, phrases,
symbols, or other meaningful elements called tokens.
• Problem? ‫فسيكفيكهم‬
• Tools? NLTK (Pyhton), CLTK, Pyarabic
• Farasa Segmenter.
Arabic Question Answering
• Open-ended QA systems.
• Data sets.
• Large Scale Retriever.
• Machine Reading Comprehension.
References
• https://medium.com/@ageitgey/machine-learning-is-fun-part-5-
language-translation-with-deep-learning-and-the-magic-of-
sequences-2ace0acca0aa
• https://medium.com/@ageitgey/machine-learning-is-fun-
80ea3ec3c471
• https://www.quora.com/What-is-machine-learning-4
• https://www.superdatascience.com/machine-learning/
• https://emerj.com/ai-glossary-terms/what-is-machine-learning/
• https://arxiv.org/pdf/1807.10854
Want to Start!
• Machine Learning A to Z Course English Version.
• Machine Learning A to Z Course Arabic Version.
• Machine Learning Course by Andrew NG.
• Machine Learning Nanodegree by Udacity NTL scholarship.
• Tensor Flow from A to Z by Andrew NG.
Arabic Natural Language Processing
• Challenges:
• Used by Hundreds of Millions of People.
• The language’s complex nature (derivational, inflectional,
etc.)
• The existence of many variations of the
language such as:
• the Classical Arabic (CA)
• (which is mainly used in
ancient and theological texts but is still understood due to its use
in the Holy Quran),
• the Modern Standard Arabic (MSA)
• (which is a modernized and simplified version of CA).
• Dialectal Arabic (DA)
• (where each region has its own dialect)

You might also like