Scribd
Upload
72 views

M. Raihan

The document discusses the limitations of perceptrons and how multilayer perceptrons can solve nonlinear problems. It explains that while a single perceptron can only learn linearly separable problems, multilayer perceptrons using backpropagation can learn any Boolean function. It then discusses decision boundaries and feature vectors in pattern recognition. Finally, it provides an overview of multilayer perceptrons and how they can learn nonlinear problems using multiple layers of nodes between the input and output layers.

Uploaded by

Sabuj Dhali
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
72 views

M. Raihan

The document discusses the limitations of perceptrons and how multilayer perceptrons can solve nonlinear problems. It explains that while a single perceptron can only learn linearly separable problems, multilayer perceptrons using backpropagation can learn any Boolean function. It then discusses decision boundaries and feature vectors in pattern recognition. Finally, it provides an overview of multilayer perceptrons and how they can learn nonlinear problems using multiple layers of nodes between the input and output layers.

Uploaded by

Sabuj Dhali
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 18

M.

Raihan
Email: [email protected]
Perceptron
Limitations of the Perceptron

25-Sep-17 3
Perceptron Limitations
Multi-layer perceptron can solve this problem
More than one layer of perceptrons (with a hard-
limiting activation function) can learn any Boolean
function
However, a learning algorithm for multi-layer
perceptrons has not been developed until much later
Back propagation algorithm (replacing the hard-limiter with a sigmoid activation
function)

25-Sep-17 4
Perceptron Limitations

25-Sep-17 5
Decision Boundaries

25-Sep-17 6
Decision Boundaries

25-Sep-17 7
Decision Boundaries
 In general, a pattern classifier carves up (partitions) the feature
space into volumes called decision regions.
All feature vectors in a decision region are assigned to the same
category.
The decision regions are often simply connected, but they can be
multiply connected as well, consisting of two or more non-touching
regions.
In pattern recognition and machine learning, a feature vector is an n-
dimensional vector of numerical features that represent some object.
Many algorithms in machine learning require a numerical
representation of objects, since such representations facilitate
processing and statistical analysis. 25-Sep-17 8
The Multi-Layer Perceptron

25-Sep-17 9
The Multi-Layer Perceptron
A multilayer perceptron (MLP) is a feed-forward artificial
neural network that generates a set of outputs from a set of
inputs.
An MLP is characterized by several layers of input nodes
connected as a directed graph between the input and
output layers.
MLP uses backpropogation for training the network.
MLP is a deep learning method.

25-Sep-17 10
MLP Decision Boundary – Nonlinear
Problems, Solved!

25-Sep-17 11
Machine Learning
What is Learning?
Learning takes place as a result of interaction between an
agent and the world.
The idea behind learning is that
Percepts received by an agent should be used not only for
acting, but also for improving the agent’s ability to behave
optimally in the future to achieve the goal.

25-Sep-17 12
Why “Learn” ?
Learning is used when:
Human expertise does not exist (navigating on Mars)
Humans are unable to explain their expertise(speech
recognition)
Solution changes in time (routing on a computer network)
Solution needs to be adapted to particular cases (user
biometrics)

25-Sep-17 13
Things You Might Be Interested In

25-Sep-17 14
Things You Might Be Interested In

25-Sep-17 15
Learning from Data
The world is driven by data.
Germany’s climate research centre generates 10 petabytes per year
Google processes 24 petabytes per day
The Large Hadron Collider produces 60 gigabytes per minute (~12 DVDs)
There are over 50m credit card transactions a day in the US alone.

25-Sep-17 16
If Data Had mass, The Earth Would Be
A Black Hole
Around the world, computers capture and store
terabytes of data everyday.
Science has also taken advantage of the ability of
computers to store massive amount of data.
The size and complexity of these data sets means that
humans are unable to extract useful information from
them.

25-Sep-17 17
Thank You

25-Sep-17 18

You might also like