Scribd
Upload
7 views7 pages

Deep Learning

Uploaded by

debbytsang1009
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
7 views7 pages

Deep Learning

Uploaded by

debbytsang1009
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/ 7

Lecture #9

Part I: What is Deep Learning?


- Neural Network & Neurons
- Activation Functions
- Layers
- Features in Layers
- Popular Neural Network Architectures
- Deep Learning vs. Traditional Machine Learning
Part II: Applications in Deep Learning
What is Deep Learning?
ChatGPT, Apple Live Text, Amazon Alexa, Tesla Autopilot,
…

Source of human’s intelligence: OUR


BRAIN

Why don’t we simulate how our brain


works?

Deep Learning mimics human’s brains.


- build an artificial neural
network
- organize neurons in layers
HOW many
- Deep: TO DESIGN?
layers
Layers and Neurons: Designed by
yourself.
Basic Units: Neurons
A Neuron in BRAIN A Neuron in DL

1. Receives ELECTRICAL 1. Receives NUMBERS


SINGALS from other from other Neurons
Neurons (Dendrites)

2. Generates an 2. Calculates a NUMBER


ELECTRICAL SIGNAL based on received
based on received numbers
signals
3. Passes the generated 3. Passes the calculated
ELECTRICAL SIGNAL to NUMBER to others as an
others input

f(x): Activation Function Activation Function makes the


- Process the outputs from the previous neurons
decision based on weighted
b: bias sum of inputs (and bias).
- An INDEPENDENT input
- for better model training
Activation Functions:
Decision Maker
A neuron uses an Activation Function to decide what to output.
A decision made by THIS NEURON

Some frequently-used Activation Functions:

Sigmoid tanh ReLU

Relatively Popular
Neurons are organized in
layers
In DL, neurons are organized in layers.

Each neuron is connected to ALL


neurons in the next layer.

Fully Connected Layer.

Input Layer: The input of the model, e.g., pictures and text
Output Layer: The last layer whose outputs are the model results
Hidden Layers: All the layers in between
ow DEEP? -> Depth: [The number of hidden layers] + [Output Layer]
It does NOT have neurons in Input Layer.
Only count the layers WITH NEURONS
Layer size: The number of NEURONS in this layer.
- Layer 1: 3
- Input Layer: 4
(The size of inputs. We may ASSUME that the neurons exist.)
An Example: Handwriting
Recognize digits from images.
- Input: An image of a digit
- Output: What the digit is
An image consists of several points
in TABLE.

In computers,
- Points: pixels
- A pixel has a value, indicating its
grey shade. (Black-White)
Probabilities
of all categories

9 with the
highest
probability

https://www.3blue1brown.com/lessons/neural-networks
Layers Learn Different
Features
Learning Hierarchical Representation:
- A Neural Network uses a MULTI-STAGE way to learn our provided data.
Small parts Larger parts Overall faces

You might also like