Handwriting Recognition

• Search Engine: Web page ranking

• Collaborative Filtering: Amazon: online bookstore

• Automatic Translation of Documents:

learn to translate a text to understand

• Security Applications: Face recognition

• Back propagation

TYPES OF MACHINE LEARNING

Supervised Learning

• Algorithms are trained using examples such as an input where the desired output is known.

Unsupervised Learning

• Used against data that has no historical labels. The system is not told the "right answer".

major- II

MACHINE LEARNING

WHY MACHINE LEARNING

• High-value predictions that can guide better decisions and smart actions in real time without human intervention
• Analyze bigger, more complex data and deliver faster, more accurate results even on a very large scale
• Where Human expertise does not exist (navigating on Mars)

APPLICATIONS

• Machine learning is programming computers to optimize a performance criterion using example data or past experience
• Study of algorithms that improve their performance using example data or past experience

Future work

• Results are expected to be more accurate.
• A bigger database should be built later to make machine learn more.
• Image with higher pixel size can be analyzed in future.
• Alphabets can be recognized using this algorithm as like numeric values.

Types of Handwriting

recognition

Offline

• Conversion of text in an image into letter codes which are usable within computer

Online

• Text as it is written on a special digitizer or PDA

Results

Handwriting Recognition

Data is loaded and

analyzed

Applications

Ability of a computer to receive and interpret intelligible handwritten input from sources such as paper documents, photographs, touch screens and other devices.

Banking purpose

• Banking applications, government documents

Signature in Retail

• Financial transactions, crossing international borders.

Some examples

Handwriting Recognition

(Using Machine learning )

Advantage & Disadvantage

ADVANTAGES

• Security Purpose bank cheque, postal code verification
• Good for those who can't use the keyboard

DISADVANTAGES

• High Variability, Writing Varies frequently.
• Hacking account by having a similarly accurate style of writing

Layout Of GUI

Submitted by

Back Propagation

• Sigmoid gradient

Neural Network

• Random initialization

IMPLEMENTATION

Typically organized in layers ,made up of a number of interconnected 'nodes' which contain an 'activation function'.

Limitations

• Gradient checking

Difference from Conventional

Computing

Neural Network

• Regularized Neural Networks

Back Propagation

• Slower to train than other types of networks and require thousands of Data
• User has no other role than to feed it input and watch it train and await the output. learning itself progresses on its own.

• No complex central processors like conventional computers
• ANN do not execute programed instructions, respond in parallel to the pattern of inputs presented to it.
• No separate memory addresses for storing data for ANN.

Ashish verma

Visualizing the hidden layer

2k12/ep/020

Neural Network

• Visualizing the data

Uses

• Model representation

 Capturing associations or discovering regularities within a set of patterns where the volume, number of variables or diversity of the data is very great

 The relationships between variables are vaguely understood. The relationships are difficult to describe adequately with conventional approaches.

• Feed forward & cost function

• Regularized cost function