Scribd
Upload
7 views22 pages

Introduction To Machine Learning

The workshop on Machine Learning aims to cover its definition, classifications, principles, and popular models, along with practical tools like Scikit-Learn. It highlights the significance of machine learning in processing vast amounts of data and solving various problems through supervised, unsupervised, and reinforcement learning. Additionally, it introduces the workflow for machine learning projects and tools like Jupyter Notebook and Google Colab for implementation.

Uploaded by

rolypoly
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
7 views22 pages

Introduction To Machine Learning

The workshop on Machine Learning aims to cover its definition, classifications, principles, and popular models, along with practical tools like Scikit-Learn. It highlights the significance of machine learning in processing vast amounts of data and solving various problems through supervised, unsupervised, and reinforcement learning. Additionally, it introduces the workflow for machine learning projects and tools like Jupyter Notebook and Google Colab for implementation.

Uploaded by

rolypoly
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 22

Introduction

to Machine
Learning
LEI QIAN, PH.D., MEHARRY MEDICAL
COLLEGE
NSF NDS A WORKSHOP, OCT 29-30, 2024
The Goals of This
Workshop
1. What is Machine Learning - its relation to AI and Deep
Learning
2. Classifications of Machine Learning
3. Principles and key components of Machine Learning -
bias, variance, training, testing, and validation data
etc.
4. Popular Machine Learning models: Neural Network,
Support Vector Machine, Decision Tree, Random
Forest, etc.
5. Sciket-Learn - a powerful and popular ML library.
What is Machine
Learning
◦ Machine learning is a subfield of Aritifical Intelligence (AI) that explores the
study and construction of algorithms that can learn from and make decisions
on data without having to program.
◦ Instead of relying on predefined rules, ML algorithms use patterns in data to
improve performance over time.

◦ Examples of ML:
◦ Speech Recognition
◦ Object detection
◦ Playing games
◦ Medical diagnosis
◦ Anomaly detection
◦ GenAI (ChatGPT, Copilot, etc)
What is Artificial
Intelligence
◦ In computer science, artificial intelligence (AI), sometimes called machine
intelligence, is intelligence demonstrated by machines. It is a broader concept
comparing to Machine Learning.
◦ The term "artificial intelligence" is often used to describe machines (or
computers) that mimic "cognitive" functions that humans associate with the
human mind, such as "learning" and "problem solving"
◦ Challenges of AI includes reasoning, problem solving, knowledge
presentation, etc.
◦ Methods of AI includes symbolic reasoning, statistical learning etc.
What is Artificial
Intelligence

https://hackernoon.com/difference-between-artificial-intelligence-ma
chine-learning-and-deep-learning-1pcv3zeg
Why machine learning:
We are drowning in information and starving for knowledge. — John
Naisbitt
◦ 500 hours of video are uploaded to YouTube every minute, amount to one
year of video every 18 minutes of uploaded! (2022)
◦ Google stores 10 exabytes of data, which is bytes of data!
◦ Facebook process 500 TB of data each day!
◦ Each human genome has a length of base pairs. Genomes of thousands of
peoples have been sequenced.
Get knowledge from this immense amount of data requires computer and
machine learning algorithms.
What Problems can be
Solved by Machine
Learning?
1. Machine learning needs data to train models.
•Text data
•Image data
•Audio data
•Other numerical data (length, age, rating, price, etc)
•Categorical data (gender, positive/negative, shape, types etc)
2. We can apply machine learning only when data has pattern.
3. If the problem has explicit mathematics formulas, we do not need
machine learning.
Types of machine
learning
Machine learning is usually divided into three types:

1. Supervised learning
2. Unsupervised learning
3. Reinforcement learning.

Supervised learning is by far the most widely used.

Other by-product of machine learning:


◦ Generative models (GenAI)
◦ Autoencoder
Supervised Learning
◦ Supervised Learning
◦ The machine learning task of inferring a function from labeled training data.
◦ A set of labeled data is used to train the system.
◦ The trained system can be applied to test data to make prediction.
◦ Examples:
◦ Recognizing objects
◦ Identifying spam emails
◦ Predicting stock prices
◦ Natural language recognition
◦ Protein structure prediction
◦ Children adult height prediction
Supervised Learning
◦ Supervised Learning can be categorized into regression and
classification
◦ Regression: the dependent variable is a real value variable.
◦ Stock price prediction
◦ Children's adult height prediction
◦ Weather temperature prediction
◦ Sale forcast
◦ Classification: the dependent variable is a categorical variable
◦ Classify an email to SPAM or non-SPAM (2-classes).
◦ Identify what objects are in a picture (multiple classes)
◦ Disease diagnosis (2-classes)
Unsupervised Learning
◦ Unsupervised Learning
◦ the goal is to find “interesting patterns” in the data.
◦ The data has no target attribute.
◦ Data is not labeled.
◦ No obvious error metric to use (unlike supervised learning).
◦ Sometimes called knowledge discovery.
◦ Examples:
◦ Image clustering
◦ Anomaly detection
◦ Gene Sequence Analysis
Reinforcement Learning
◦ Reinforcement learning: an agent learns to make decisions by
interacting with an environment. It is inspired by the way humans
and animals learn from trial and error to achieve certain
objectives.
◦ Examples:
◦ Game playing
◦ Robotics
◦ Autonomous vehicles
◦ Natural Language Procesing
Large Language Models
Large Language Models such as ChatGPT, Gemini, MS Copilot,
Github Copilot, Llama, etc are trained by a combination of
unsupervised, supervised, and reinforcement learning in difference
phases.

◦ In the Pretraining stage, unsupervised or self-supervised.


◦ In the Fine-Tuning stage, it is supervised.
◦ After that, the model can be training by reinforcement learning
from Human Feedback (RLHF)
Machine Learning
Workflow
1. Problem definition
2. Data Collection
3. Data Preprocessing
4. Model Selection
5. Model Training/Validation
6. Model Evaluation
7. Final Deployment
8. Monitoring and Maintenance
Jupyter Notebook
◦ Jupyter Notebook is an open-source web application that allows
you to create and share documents containing Python code and
Markdown text.
◦ It is considered the de facto standard for many machine learning
(ML) development workflows, thanks to its flexibility,
interactivity, and ease of use.
◦ Jupyter Notebook Code can run:
◦ on Jupyter Notebook Server as part of Anaconda
◦ Remote servers
◦ Free online service:
◦ Google Colab colab.research.google.com
◦ Kaggle Kernels: www.Kaggle.com/kernels
◦ Microsoft Azure: notebooks.azure.com
Working with Notebook
◦ A Notebook consists of multiple cells. A cell can be:
◦ Python Code
◦ Markdown Text

The Notebooks in Google Colab is slightly different to Jupyter Notebooks


coming with Anaconda.
Working with Google
Colab
1. Use any web browser such as Chrome, MS Edge, Apple Safari, etc to
open https://colab.research.google.com
2. Click File->Open, you will be able to open a Jupyter Notebook file.
3. You can open a notebook through:
1. Google Drive (in your personal account's Colab Notebooks folder).
2. From Github
3. Upload notebook files.

You might also like