Chapter_1

introducing deep
learning
 Why you
should learn
deep learning
 Deep learning is today’s chapter in this story of
innovation

It’s A Powerful Tool For

The Incremental This is particularly extraordinary given that deep
learning seems to use largely the same brain-
Automation Of inspired algorithm for achieving these
Intelligence accomplishments across a vast number of fields

Even though deep learning is still an actively

developing field with many challenges, recent
developments have lead to tremendous
excitement: perhaps we’ve discovered not just a
great tool, but a window into our own minds
 There’s a substantial amount of hype
around the potential impacts of deep
learning if the current trend of progress is
extrapolated at varying speeds

Deep Learning Has The

Call- center operators, taxi drivers, and low-
Potential For Significant level business analysts are compelling
Automation Of Skilled examples where deep learning can
Labor provide a low-cost alternative

It’s my hope that you will be enabled by

this book to transition from perhaps one of
the industries facing disruption into an
industry ripe with growth and prosperity:
deep learning
 Personally, I got into deep learning
because it’s fascinating

It’s Fun And Creative. You’ll

Discover Much About What It Is Unpacking exactly what it means
To Be Human By Trying To to think, to reason, and to create is
Simulate Intelligence And enlightening, engaging, and, for
Creativity me, inspiring

Consider having a dataset filled

with every painting ever painted,
and then using that to teach a
machine how to paint like Monet
 There are two major groups of deep learning
educational material

One group is focused around how to use popular

frameworks and code libraries like Torch,
It will help you TensorFlow, Keras, and others
understand what’s
inside a framework
The other group is focused around teaching deep
learning itself, otherwise known as the science
under the hood of these major frameworks

This book is about teaching you what deep learning

is so you can then be prepared to learn a
framework
What you need to
get started
• My absolute favorite place to
work is in Jupyter Notebook
• One of the most important parts
of learning deep learning is the
ability to stop a network while it’s
training and tear apart absolutely
every piece to see what it looks
like
• This book teaches deep learning
from absolute scratch, soup to
nuts
Pass high school
mathematics
• Some mathematical assumptions
are out of depth for this book, but
my goal is to teach deep learning
assuming that you understand
only basic algebra
Find a personal
problem you’re
interested in
• Learning deep learning was just a
“dependency” to solving some other
interesting task
• And as it turns out, this field is so new,
and is changing so fast, that if you
spend the next couple of years chasing
one project with these tools, you’ll find
yourself becoming one of the leading
experts in that particular problem faster
than you might think
• For deep learning, having a problem
you’re fascinated with that involves
using one dataset to predict another is
the key catalyst!
You’ll probably
need some Python
knowledge

• Python is an amazingly intuitive

language
• Furthermore, the Python community
has a passion for simplicity that
can’t be beat
Summary

• If you’ve got a Jupyter notebook in

hand and feel comfortable with the
basics of Python, you’re ready for
the next chapter!
• As a heads-up, chapter 2 is the last
chapter that will be mostly
dialogue based
• It’s designed to give you an
awareness of the high-level
vocabulary, concepts, and fields in
artificial intelligence, machine
learning, and, most important,
deep learning
 Chapter 2
Fundamental
Concepts
WHAT ARE DEEP LEARNING, WHAT ARE PARAMETRIC WHAT ARE SUPERVISED HOW CAN MACHINES
MACHINE LEARNING, AND MODELS AND LEARNING AND LEARN?
ARTIFICIAL INTELLIGENCE? NONPARAMETRIC MODELS? UNSUPERVISED LEARNING?
 What Is Deep Learning?
• Deep learning is a subset of machine
learning, which is a field dedicated to the
study and development of machines that
can learn
• In industry, deep learning is used to solve
practical tasks in a variety of fields such as
computer vision , natural language
processing , and automatic speech
recognition
• In short, deep learning is a subset of
methods in the machine learning toolbox,
primarily using artificial neural networks,
which are a class of algorithm loosely
inspired by the human brain
input datasets in bold and
output datasets in italic
• Using the pixels of an image to detect the
presence or absence of a cat
• Using the movies you’ve liked to predict more
movies you may like
• Using someone’s words to predict whether
they’re happy or sad
• Using weather sensor data to predict the
probability of rain
• Using car engine sensors to predict the
optimal tuning settings
• Using news data to predict tomorrow’s stock
price
• Using a raw audio file to predict a transcript
of the audio
 Supervised learning is a method for
transforming one dataset into another

Supervised
Supervised machine learning is the
machine bread and butter of applied artificial
learning intelligence

This allows supervised machine

learning algorithms to extend human
intelligence and capabilities in a
seemingly endless number of ways
Unsupervised
machine learning
• Unsupervised learning shares a property in
common with supervised learning: it
transforms one dataset into another
• But the dataset that it transforms into is not
previously known or understood
• Unlike supervised learning, there is no
“right answer” that you’re trying to get the
model to duplicate
• Clustering transforms a sequence of
datapoints into a sequence of cluster
labels
• This idea of clustering is something you
can reliably hold onto in your mind as the
definition of unsupervised learning
 • An algorithm is either unsupervised or supervised, and
either parametric or nonparametric. Whereas the previous
section on supervision is about the type of pattern being
learned, parametricism is about the way the learning is
Parametric vs. stored and often, by extension, the method for learning.
nonparametric
learning • A parametric model is characterized by having a fixed
number of parameters, whereas

• a nonparametric model’s number of parameters is infinite

(determined by data).
Supervised
parametric learning
• Supervised parametric learning
machines are machines with a fixed
number of knobs , wherein learning
occurs by turning the knobs
• Input data comes in, is processed
based on the angle of the knobs, and is
transformed into a prediction
• Learning is accomplished by turning the
knobs to different angles
• If you’re trying to predict the probability
that the Red Sox will win the World
Series, then this model would first take
data and make a prediction
 Supervised Parametric
Learning Steps
1. Data Collection: Collect data that is relevant to the problem
you want to solve.
2. Data Preprocessing: Clean and preprocess the data to
remove any noise or inconsistencies.
3. Feature Selection: Select the most relevant features that will
help in making accurate predictions.
4. Model Selection: Choose a parametric model that is
appropriate for your problem.
5. Training: Train the model on the labeled data.
6. Validation: Validate the model on a separate set of labeled
data to ensure that it generalizes well.
7. Testing: Test the model on a separate set of labeled data to
evaluate its performance.
 Supervised Parametric
Learning Example
Suppose you want to build a model that
predicts whether a customer will buy a product
or not based on their age and income. You
collect data on 1000 customers, preprocess it,
and select age and income as relevant features.
You choose logistic regression as your
parametric model, train it on 800 labeled
examples, validate it on 100 examples, and test
it on 100 examples. Your model achieves an
accuracy of 85% on the test set
Unsupervised
parametric
learning
GROUP MEMBERSHIP
PROBABILITY
Nonparametric
learning
• Nonparametric learning is a class of algorithm
wherein the number of parameters is based on data
• This lends itself to methods that generally count in
one way or another, thus increasing the number of
parameters based on the number of items being
counted within the data
• What makes this simple model nonparametric is this
trait wherein the number of parameters changes
based on the data
• This is in contrast to parametric models, which start
with a set number of parameters and, more
important, can have more or fewer parameters
purely at the discretion of the scientist training the
model
• Even parametric algorithms are somewhat
influenced by the number of classes in the data,
even if they aren’t explicitly counting patterns
Summary

• In this chapter, we’ve gone a level

deeper into the various flavors of
machine learning
• You learned that supervised
machine learning is a class of
algorithm where you learn to
predict one dataset given another
and that unsupervised learning
generally groups a single dataset
into various kinds of clusters
• Deep learning uses neural networks
to perform both supervised and
unsupervised prediction