What is Machine Learning
LONDON INNOVATION ACADEMY 1
What is Machine Learning?
Machine Learning is a branch of artificial intelligence (AI) that enables machines to learn from data without being
explicitly programmed for each task. Instead, models are provided with data, allowing them to find patterns or
relationships, which they can then apply to make predictions or decisions on new data.
LONDON INNOVATION ACADEMY 2
Types of Machine Learning and Their Classification
We can classify machine learning into several types based on how models learn from data:
1. Supervised Learning
In supervised learning, the model is trained on data that includes input (features) and output (target) pairs. In other
words, the model knows the expected outcome and adjusts itself based on that data to reach the correct answer.
• Examples:
o Classification: The model categorizes inputs into specific groups. For example, predicting whether an email
is "spam" or "not spam" based on the email's content.
o Regression: The model predicts a continuous value based on input features, such as predicting house prices
based on square footage and number of rooms.
LONDON INNOVATION ACADEMY 3
2. Unsupervised Learning
In unsupervised learning, the model is trained only on input data without any known outcomes. The goal is to
discover patterns or structures within the data. This type is often used for data where we don’t have a clear idea of the
target outcome.
• Examples:
o Clustering: The model groups data into clusters based on similarity. An example is customer segmentation,
where customers are grouped based on purchasing behavior.
o Principal Component Analysis (PCA): This technique reduces data dimensions, making it easier to analyze
and visualize while preserving as much information as possible.
LONDON INNOVATION ACADEMY 4
3. Reinforcement Learning
In reinforcement learning, the model learns by interacting with an environment, receiving "rewards" or "penalties"
based on its actions. This approach is ideal for applications that require continuous, sequential decision-making, such
as in gaming or autonomous vehicles.
• Examples:
o Deep Learning in Gaming: The algorithm learns how to play a game and improve its performance based on
the rewards received for each move.
o Robotics: A robot learns to navigate through a factory and achieve specific tasks, learning through trial and
error.
LONDON INNOVATION ACADEMY 5
4. Semi-Supervised Learning
Semi-supervised learning combines supervised and unsupervised learning. The model is trained on a dataset that
includes a small number of labeled examples and a large number of unlabeled examples. This is useful in situations
where labeled data is expensive or scarce.
• Examples:
o Image Recognition: The model is provided with a few labeled images (e.g., images of people with their
names) and many unlabeled images, allowing it to learn from both.
5. Self-Supervised Learning
Self-supervised learning is a newer type of machine learning, where the system learns complex patterns independently
from the data. It extracts internal relationships between parts of the data without human intervention.
• Examples:
o Image and Text Processing in Deep Learning: For example, generating captions for images without prior
labeling, often using neural networks to learn intricate patterns within the dataset.
LONDON INNOVATION ACADEMY 6
Famous Examples of Machine Learning Applications by Type
1. Supervised Learning:
o Linear Regression: Used for predicting numerical values.
o Random Forest: Can be used for classification or
regression.
2. Unsupervised Learning:
o K-Means Clustering: Used to group data into clusters.
o Principal Component Analysis (PCA): Used for
dimensionality reduction and pattern analysis.
3. Reinforcement Learning:
o Q-Learning Algorithm: Commonly used in robotics and
video games where the algorithm must make sequential
decisions.
4. Semi-Supervised Learning:
o Classification Algorithms such as SVM, improved with
unlabeled data for better accuracy.
5. Self-Supervised Learning:
o Generative Models like GANs (Generative Adversarial Networks) that generate new images or data based on relationships within the
available data.
LONDON INNOVATION ACADEMY 7
here’s a structured and translated
breakdown of the concepts covered,
incorporating insights from the Hands-On
Machine Learning with Scikit-Learn, Keras,
and TensorFlow book where relevant. I’ll
keep this organized into sections for clarity.
LONDON INNOVATION ACADEMY 8
How Does Machine Learning Work?
• The inner workings of machine learning can be broken down into three key steps:
Data Acquisition and Preparation
• The foundation of any successful ML project is data. The quality and quantity of data significantly impact
the model’s performance. Data can come from various sources like sensors, social media platforms,
customer transactions, or even scientific experiments.
• However, raw data is often messy and needs cleaning, organizing, and formatting for the model to
understand it.
LONDON INNOVATION ACADEMY 9
Model Training
• Once the data is prepared, it’s fed into the chosen machine learning algorithm. This algorithm acts like a
mathematical formula that learns from the data.
• Different algorithms are suited for different tasks. For example, a Support Vector Machine (SVM) might be
used for image classification, while a Recurrent Neural Network (RNN) might be better for tasks involving
sequential data like language translation.
• During training, the model adjusts its internal parameters to minimize errors in its predictions. Imagine a
student fine-tuning their study approach based on exam results – that’s the essence of model training.
LONDON INNOVATION ACADEMY 10
Evaluation and Deployment
• Once trained, the model’s performance is evaluated on a separate dataset to assess its accuracy and
generalizability. If the model performs well, it can be deployed for real-world applications.
• This could involve integrating it into a software program, a mobile app, or even a physical device like a self-
driving car.
LONDON INNOVATION ACADEMY 11
