INTRODUCTIONTO DEEP LEARNING
Felicia O’Garro | Code Crew Meetup | February 4, 2021
WHAT IS DEEP LEARNING?
• Technique that can be used to implement Machine Learnin
g

• Analyzes and learns from data to make prediction
s

• Learning can be supervised, semi-supervised and unsupervised
ARTIFICIAL NEURAL NETWORK
• Algorithms inspired by the structure and function of the brain’s
neural networ
k

• Consist of connected units called arti
fi
cial neurons that can transmit
signals to each other
ARTIFICIAL NEURAL NETWORK CON’T
TRAIN/TEST SPLIT
• Technique for evaluating performance
 

• Training se
t

• Validation se
t

• Testing set
LAYERS
• Building blocks of a neural networ
k

• Extract representations from the data
fed into layers
ACTIVATION FUNCTIONS
• Sigmoid - transforms input to a value
between 0 and 1
 

• Hyperbolic tangent (Tanh) - transforms
input to a value between -1 and
1

• ReLu - transforms input to a max of 0
or the input itself
GRADIENT DESCENT
LOSS FUNCTIONS
• Sparse Categorical Cross-Entrop
y

• Binary Cross-Entrop
y

• Mean Squared Erro
r

• Mean Absolute Erro
r

• And more…
EPOCHS
PREDICTIONS
CODE WALKTHROUGH
REAL WORLD USE CASES
• Image Classi
fi
catio
n

• Facial Recognition
 

• Self Driving Cars
 

• Art Creation
 

• Music Compositio
n

• Fraud Detectio
n

• Game
s

• Chatbot
s

• Clothing Designs
RESOURCES
• PythonTutorial 2020
• Khan Academy Linear Algebra
 

• fast.ai
• Kaggle
 

• Google Colab
• PyTorch
 

• Tensor
fl
ow
• Coursera Deep Learning Specialization
• Deep Lizard
• Hands-on Machine Learning with Scikit-Learn, Keras &Tensor
fl
ow
THANKS MUCH!!!

Introduction to Deep Learning

  • 1.
    INTRODUCTIONTO DEEP LEARNING FeliciaO’Garro | Code Crew Meetup | February 4, 2021
  • 2.
    WHAT IS DEEPLEARNING? • Technique that can be used to implement Machine Learnin g • Analyzes and learns from data to make prediction s • Learning can be supervised, semi-supervised and unsupervised
  • 3.
    ARTIFICIAL NEURAL NETWORK •Algorithms inspired by the structure and function of the brain’s neural networ k • Consist of connected units called arti fi cial neurons that can transmit signals to each other
  • 4.
  • 5.
    TRAIN/TEST SPLIT • Techniquefor evaluating performance • Training se t • Validation se t • Testing set
  • 6.
    LAYERS • Building blocksof a neural networ k • Extract representations from the data fed into layers
  • 7.
    ACTIVATION FUNCTIONS • Sigmoid- transforms input to a value between 0 and 1 • Hyperbolic tangent (Tanh) - transforms input to a value between -1 and 1 • ReLu - transforms input to a max of 0 or the input itself
  • 8.
  • 9.
    LOSS FUNCTIONS • SparseCategorical Cross-Entrop y • Binary Cross-Entrop y • Mean Squared Erro r • Mean Absolute Erro r • And more…
  • 10.
  • 11.
  • 12.
  • 13.
    REAL WORLD USECASES • Image Classi fi catio n • Facial Recognition • Self Driving Cars • Art Creation • Music Compositio n • Fraud Detectio n • Game s • Chatbot s • Clothing Designs
  • 14.
    RESOURCES • PythonTutorial 2020 •Khan Academy Linear Algebra • fast.ai • Kaggle • Google Colab • PyTorch • Tensor fl ow • Coursera Deep Learning Specialization • Deep Lizard • Hands-on Machine Learning with Scikit-Learn, Keras &Tensor fl ow
  • 16.