Scribd
Upload
62 views

Machine Learning

This document provides an overview of machine learning and deep learning. It discusses how machine learning evolved from symbolic AI by allowing computers to learn from examples without being explicitly programmed. Machine learning works by training models on large datasets to find statistical patterns and make predictions. Deep learning is a modern approach that uses multiple layers of representation to learn directly from raw data in an end-to-end fashion. Examples of deep learning applications include image classification, speech recognition, and machine translation.

Uploaded by

Phani Karthik Maradani
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
62 views

Machine Learning

This document provides an overview of machine learning and deep learning. It discusses how machine learning evolved from symbolic AI by allowing computers to learn from examples without being explicitly programmed. Machine learning works by training models on large datasets to find statistical patterns and make predictions. Deep learning is a modern approach that uses multiple layers of representation to learn directly from raw data in an end-to-end fashion. Examples of deep learning applications include image classification, speech recognition, and machine translation.

Uploaded by

Phani Karthik Maradani
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

Machine Learning

Phani Karthik Maradani


AI , ML & DL

Artificial
Intelligence

Machine Learning

Deep Learning
Artificial Intelligence
• Born in 1950s
– “Whether computers could be made to think ?” – still exploring the ramifications
– The effort to automate intellectual tasks normally performed by Humans

• Symbolic AI (1950s – late 1980s)


– Eg: Early chess program based on hardcoded rules crafted by programmers – doesn’t
qualify as ML
– Suitable to solve well-defined, logical problems
– Intractable for complex, fuzzy problems like image classification, speech recognition
and language translation
Machine Learning
• Can computer go beyond “what we know
how to order it to perform” and learn on
its own how to perform a specified task” ?
• Could a computer surprise us ?
• Rather than programmers crafting
data-processing rules by hand, could a
computer automatically learn these rules
by looking at data ?
New Programming
Paradigm

Rules Classical
Answers
Data programming

Data
Machine Learning Rules
Answers
Machine Learning
• Trained rather than explicitly programmed
• Presented with relevant examples to a task
• Finds statistical structure in these
examples
• Tagging vacation pictures
• Started to flourish in 1990s, on availability
of faster hardware and larger datasets
• Hands-on discipline – ideas proven
empirically than theoretically
Three things needed for ML
• Input data points
– Speech recognition – sound files
– Image recognition - pictures

• Examples of expected output


– Transcripts of sound files
– Tagged images

• A way to measure whether algorithm is


doing good job
– Used as a feedback signal to adjust the way the algorithm works
Learning representations from Data
• Central problem is to meaningfully
transform data
• Learn useful representations of input
data – representations that get us
closer to expected output
• Different way to look at data (represent
or encode)
– Color image (RGB vs HSV – Select all red pixels vs reduce saturation)
y
Binary Classification
Develop an algorithm given ( x, y ) can output
whether that point is likely to be black or white

Better Representation
Raw Data Coordinate Change
y
y x
y

x
x
Deep Learning
• Emphasis on learning successive layers of increasingly
meaningful representations
• Inspired by the understanding of the brain – not models of the
brain – No evidence
• Deep learning is a mathematical framework for learning
representations from data
Input X

Layer
Weights
(data transformation)

Layer
Weights
(data transformation)

Predictions
True targets
Y

Loss
Optimizer
function

Loss score
Video by:
Andreas Stöckl
Video by:
Andreas Stöckl
Examples
• image classification
• speech recognition
• handwriting transcription
• machine translation
• text -to-speech conversation
• Digital assistants - Google Now and Amazon Alexa
• autonomous driving
• Ad targeting eg. Google
• Search results
• Answering bots
Questions ?
Thank
You
[email protected]

Reference: Deep Learning with Python by Francois Chollet

You might also like