You are on page 1of 16

INTRODUCTION TO

ARTIFICIAL NEURAL NETWORKING.


By

RAHUL KUMAR ( 23MDM4043 )


M.Tech Medical Devices
(2nd Semester)
What is ANN ( Artificial Neural Network ) Model

An artificial neural network is an computational model inspired by the structure and


function of the human brain ( ANN resemble the human brain in their structure and
function. They are inspired by the biological neurons signal to one another. ANN
consist of interconnected nodes, or artificial neurons, in a layered structure, similar to
the neural pathways in the human brain.)

ANNs are used for predictive modelling. ( Predictive modelling Is the process of using
data and statistical algorithms to predict outcomes. ANN are commonly used in
predictive modelling due to their ability to learn )
Neural network (neural + network)

Connection
Neuron

They consist of an input layer, one or more hidden layers, and an output layer,
with signals traveling through the network and possibly passing through multiple
intermediate layers. The network learns from experience and can derive
conclusions from complex and seemingly unrelated information. ANNs are
capable of learning and modelling, non-linearities and complex relationships,
making them a powerful tool in data analysis and pattern recognition.
INPUT DATA MEANING
Neural Networks

Algorithm

LEARNING
&
IMPROVEMENT
Dendrites ( Fetch input )

Axon ( Data Pass )

Cell body ( Data Processing )

Terminal axon
Associated weights
Inputs

X1 W1

W2
X2
f (x) Y ( Output )

W3

X3
Patterns in Artificial Neural Networks

The design of pattern recognition systems usually involves three


aspects:

Data acquisition & Preprocessing

Data Representation

Decision-making approaches
Data Acquisition:
This step involves gathering the data needed to train and test the neural network model.
The data can come from various sources such as databases, files, sensors, or APIs. It's
crucial to ensure that the data collected is relevant to the problem being solved and is of
good quality. This may involve data cleaning to remove errors, missing values, or
outliers.

Data Preprocessing:
Once the data is acquired, it often needs to be preprocessed to make it suitable for training the
ANN model. Preprocessing involves several steps:

1. Normalization/Standardization: This step scales the features to a similar range to ensure


that no single feature dominates the learning process. Normalization scales the data to a range
between 0 and 1, while standardization scales the data to have a mean of 0 and a standard
deviation of 1.
Feature Engineering: Feature engineering involves selecting, transforming, or creating new
features from the existing ones to improve the performance of the model. This step can include
techniques like dimensionality reduction (e.g., PCA), encoding categorical variables, or creating
new features based on domain knowledge.

1. Handling Missing Values: Missing values in the dataset need to be addressed before training the
model. This can involve imputation techniques such as mean/mode imputation, interpolation, or
using advanced methods like K-nearest neighbors (KNN) imputation.

2. Removing Outliers: Outliers can adversely affect the performance of the model. Outlier detection
techniques such as Z-score, IQR (Interquartile Range), or visual inspection can be used to identify
and remove outliers from the dataset.

3. Splitting Data: The dataset is typically split into training, validation, and testing sets. The training set
is used to train the model, the validation set is used to tune hyperparameters and monitor the
model's performance during training, and the testing set is used to evaluate the final performance of
the model.
Methods for Pattern recognition task

• Supervised learning
• Uses labelled data to train the network for pattern
1 recognition.

• Unsupervised Learning
• Allows the network to identify patterns from
2 unlabelled data.

• Reinforcement Learning
• Involves learning through continuous interaction
3 with the environment.
Supervised Learning
Supervised learning for pattern recognition in artificial neural networks (ANNs) involves training
the network on a labeled dataset. This means each input data point is paired with a
corresponding target output. For example, in image recognition, the input might be an image,
and the target output is the label describing what's in the image (like "cat" or "dog").

Unsupervised Learning
Unsupervised learning for pattern recognition in artificial neural networks (ANNs) is like discovering
hidden patterns in a jigsaw puzzle without any guide or picture on the box. You're given a bunch of
puzzle pieces without knowing what the final image looks like. Your task is to group similar pieces
together based on their shape, color, or other features. Through this process, you might uncover
clusters that represent different parts of the picture. Similarly, in unsupervised learning with ANNs, the
network is presented with data but without explicit labels. It's tasked with finding hidden structures or
patterns in the data on its own, such as clustering similar data points together or reducing the
dimensionality of the data to reveal underlying relationships. It's like letting the network explore and
discover the structure of the data without being told what to look for.
Reinforcement Learning
In reinforcement learning with ANNs, the network learns to make decisions and take
actions in an environment to maximize a reward signal. It interacts with its environment,
receiving feedback in the form of rewards or penalties based on its actions.
References
1) Wu, W., Dandy, G. C., & Maier, H. R. 2019, April.

2) https://www.ibm.com/topics/neural-networks

3) https://aws.amazon.com/what-is/neural-network/

4)
http://www.xenonstack.com/blog/artificial-neural-network-application
s

5)
https://www.spiceworks.com/tech/artificial-intelligence/articles/what-
is-pattern-recognition/amp/
THANKYOU

You might also like