Scribd
Upload
194 views12 pages

Machine Learning

Uploaded by

Connecto Prabhaa
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
194 views12 pages

Machine Learning

Uploaded by

Connecto Prabhaa
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/ 12

INTRODUCTION TO MACHINE

LEARNING
- UNIT 1
REVIEW OF LINEAR ALGEBRA FOR MACHINE LEARNING

1. Basics :
•Scalars: Single values, denoted as lowercase letters (e.g., aaa).
•Vectors: Ordered lists of numbers denoted as bold lowercase (e.g., v\mathbf{v}v).
•Matrices: 2D arrays of numbers denoted as bold uppercase (e.g., A\mathbf{A}A).
•Tensors: Generalizations of vectors & matrices to higher dimensions.

2. Key Operations :

•Addition/Subtraction: Element-wise for same-sized matrices or vectors.

•Scalar Multiplication: Each element is multiplied by a scalar.

•Dot Product: Multiplication of two vectors to yield a scalar (measures similarity).

•Matrix Multiplication: Combines two matrices; central for transformations


(applying weights).
INTRODUCTION AND MOTIVATION FOR MACHINE
LEARNING
Machine learning is a subset of artificial intelligence that involves using algorithms
and statistical models to enable computers to perform tasks without being explicitly
programmed. ML allows systems to learn from data and improve their performance
over time based on experience.

1. Types of Machine Learning:

a. Supervised learning – Also called predictive learning. A machine predicts the


class of unknown objects based on prior class-related information of similar
objects.

b. Unsupervised learning – Also called descriptive learning. A machine finds


patterns in unknown objects by grouping similar objects together.

c. Reinforcement learning – A machine learns to act on its own to achieve the


given goals.
EXAMPLES
1. Image Recognition and Computer Vision

•Facial Recognition: Used in social media, surveillance, and smartphone


security, facial recognition systems can identify or verify people based on their
facial features

•Object Detection: Self-driving cars rely on computer vision to detect and


classify objects like other cars, pedestrians, and traffic signals, ensuring safe
navigation.

•Medical Imaging: AI models help radiologists detect abnormalities (like


tumors) in X-rays, MRIs, and CT scans, aiding early diagnosis and treatment
planning.

•Augmented Reality (AR): Image recognition algorithms track real-world


objects, which AR applications then overlay with virtual content for gaming,
shopping, or remote assistance.
EXAMPLES
2. Natural Language Processing (NLP)
• Language Translation: Apps like Google Translate use NLP models to
translate text or speech in real-time, supporting multiple languages.

• Sentiment Analysis: Businesses use NLP to analyze customer sentiment in


reviews or social media, helping them understand consumer opinions.

• Chatbots and Virtual Assistants: Siri, Alexa, and other voice-activated


assistants use NLP to understand spoken language, answer questions, and perform
tasks.

• Content Generation and Summarization: Tools like GPT can


generate human-like text, summarize articles, or assist with creative
writing, enhancing productivity.
VC DIMENSION
VC Dimension :

(Vapnik-Chervonenkis Dimension): The VC dimension is a measure of a

model’s capacity, or its ability to classify a variety of data patterns. It’s defined as

the maximum number of points a model can shatter, meaning the model can

perfectly classify all possible labelings of those points. Higher VC dimensions imply

more complex models, potentially leading to overfitting, while lower VC dimensions

suggest simpler models that may underfit. It’s crucial for understanding a model’s

generalization ability.
PROBABLY APPROXIMATELY CORRECT (PAC) LEARNING

Probably Approximately Correct (PAC) Learning is a framework in machine learning that quantifies a
model's ability to learn from data. In PAC learning, a model is considered successful if, with high
probability (the "Probably" part), it can learn a hypothesis that is approximately correct—that is, close
enough to the true function or distribution generating the data.

Key points:
•Probably: The model will produce an accurate hypothesis with a high probability (e.g., 95%).
•Approximately Correct: The hypothesis may not be perfect, but its error is within an acceptable margin
(ε).
•Efficiency: PAC learning also considers the computational efficiency of finding this hypothesis within a
reasonable amount of data and time.
HYPOTHESIS SPACE

The hypothesis space in machine learning is the set of all possible models or
functions that a learning algorithm can choose from to fit a given dataset. It
includes every potential hypothesis (or function) that could map inputs to outputs
based on the training data.
Why It’s Essential
•Defines Learning Scope: The hypothesis space determines the complexity and
flexibility of the models, influencing what patterns or relationships the model can
learn from the data.
•Affects Generalization: A too-large hypothesis space can lead to overfitting,
where the model learns noise instead of patterns. A too-small hypothesis space may
underfit, missing important data relationships.
•Guides Model Selection: Choosing an appropriate hypothesis space helps
INDUCTIVE BIAS

Inductive bias is the set of assumptions a machine learning model makes to


predict outputs for unseen inputs based on its training data. It provides the model
with a "preference" toward certain solutions, which helps it generalize effectively
from limited data.
Purpose:
• Generalization: Inductive bias enables a model to go beyond the training data,
predicting patterns it hasn't explicitly seen.
• Guiding Hypothesis Selection: By narrowing down the hypothesis space,
inductive bias helps the model focus on likely solutions rather than exploring
every possible one.

Types of Inductive Bias:


GENERALIZATION

Generalization refers to a model's ability to make accurate predictions on new,


unseen data, based on patterns learned from the training data. It is a key goal, as a
model that generalizes well performs well not only on the training data but also on
previously unseen test data.
• Good generalization means the model can handle new data effectively.
• Overfitting occurs when the model is too specific to the training data and
performs poorly on new data.
• Underfitting occurs when the model is too simple to capture the underlying
patterns in the data.
• The bias-variance tradeoff affects generalization: models with high bias may
underfit, while those with high variance may overfit.
• Techniques like regularization help prevent overfitting and improve
BIAS-VARIANCE TRADE-OFF VISUALIZATION

Bias:
• Represents the error introduced by approximating a real-world problem, which
may be complex, with a simpler model.
• High bias models tend to make strong assumptions about the data and may
oversimplify it.
• Leads to underfitting, where the model is too simple and fails to capture the data
patterns.
Variance:
• Represents the model's sensitivity to fluctuations in the training data.
• High variance models capture noise along with the underlying data patterns.
• Leads to overfitting, where the model performs well on training data but poorly
on new, unseen data.
Trade-Off :
BIAS-VARIANCE TRADE-OFF VISUALIZATION

THANK YOU..

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy