ANn

ARTIFICIAL NEURAL NETWORK

PRESENTATION

IENG 447 CIM

Waleed Ashour ~19700694 Abdulrahman Maysarah~21008030Reem Diyab ~20911074 Mohamed el mellas~20700315

INDEX

INTRODUCTION

EXPLANATION

DEFINITION

APPLICATIONS

HISTORY

Limitations

INDEX

MAJOR CATEGORIES

01

INTRODUCTION

WHAT ARE NUERAL NETWORKS

DEFINITION

It is a type of machine learning process, called deep learning, that uses interconnected nodes or neurons in a layered structure that resembles the human brain.

ARTIFICIAL NEURAL NETWORK

NEURAL NETWORK

02

WHAT ARE NUERAL NETWORK LAYERES?

3 MAIN LAYERS

01

INPUT LAYER

--------

ARTIFICIAL NEURAL NETWORK

The input layer of a neural network is composed of artificial input neurons, and brings the initial data into the system for further processing by subsequent layers of artificial neurons.

02

HIDDEN LAYER

--------

ARTIFICIAL NEURAL NETWORK

a hidden layer is located between the input and output of the algorithm, in which the function applies weights to the inputs and directs them through an activation function as the output

03

OUTPUT LAYER

--------

ARTIFICIAL NEURAL NETWORK

The output layer is the final layer in the neural network where desired predictions are obtained.

03

SIMPLE HISTORY TIMELINE OF ANN in CIM

1950 - 2010

1950s:

Frank Rosenblatt develops the Perceptron, a simple type of ANN that can classify input data into one of two categories based on a linear combination of its features.

F. ROSENBLATT

1960s and 1970s:

Researchers develop more complex ANN architectures, such as multi-layer perceptrons and radial basis function networks, allowing ANNs to perform a wider range of tasks, including regression, clustering, and function approximation.

Bernard Widrow

Ted Hoff

1980s and 1990s:

John Hopfield- introduced the concept of energy minimization in ANNs David Rumelhart, Geoffrey Hinton, and Ronald Williams- developed the backpropagation algorithm for training multi-layer perceptrons in the 1980s. ANNs became more widely adopted in CIM applications, particularly for tasks such as process control, optimization, and fault diagnosis. However, the limitations of early ANN architectures limit their widespread use in industrial settings.

John Hopfield

Geoffrey Hinton

David Rumelhart

Ronald Williams

2000s and 2010s:

Andrew Ng: Co-founded the Google Brain project Adam Coates: Led the development of the Baidu Deep Speech system The development of new ANN architectures, such as convolutional neural networks and recurrent neural networks, as well as advances in computational hardware and software, allow ANNs to be applied to an even wider range of CIM tasks, including predictive maintenance, quality control, and supply chain management.

Andrew Ng

Adam Coates

04

MAJOR CATEGORIES OF NEURAL NETWORKS

3 MAJOR CATEGORIES

01

02

03

ARTIFICIAL NUERAL NETWORKS

RECURRENT NEURAL NETWORKS

CONVOLUTION NEURAL NETWORKS

Lorem ipsum dolor sit amet

--------

--------

--------

05

EXPLANATION

SIMPLE OWN EXPLANATION

05

APPLICATIONS

-------------------

predictive Maintenance

downtime: refers to a a priod of time when a machine, system or service isn't available or is unable to function properly.it can occur for a variety of reasons such as planned mantinance or equipment failures etc..

+INFO

Quality Control

quality control is the process of ensuring that a product meets a certain standards of quality. this typically involves inspecting and testing products to ensure that they meet specific requirements and specifications

+INFO

Production Planning

+INFO

supply chain optimization

Lorem ipsum dolor sit amet

+INFO

LIMITATIONS

Requirement for large amounts of labeled data:

• ANNs require large amounts of labeled data to train accurately.

• Lack of labeled data or difficulty obtaining it can limit ANN performance in CIM systems.

• Techniques like transfer learning and data augmentation can help make the most of available data.

Lorem ipsum dolor sit amet

LIMITATIONS

Sensitivity to noise:

• ANNs can be sensitive to noise or outliers in the data, which can negatively affect their performance.

• Noise can refer to any errors or inconsistencies in the data that may not be representative of the underlying patterns.

• In CIM systems, noise can be introduced through various sources, such as faulty sensors or human error in data collection or labeling.

• Noise can cause the ANN to learn patterns that are not representative of the true underlying relationships, leading to poor generalization performance on new, unseen data.

• It is important to preprocess the data and remove or mitigate the impact of noise in order to improve the performance of the ANN.

Lorem ipsum dolor sit amet

LIMITATIONS

Limited interpretability:

• ANNs can be difficult to interpret due to their complex structure.

• Interpretability is important for understanding the decision-making process and identifying biases or errors in the model.

• Techniques like feature visualization and saliency maps can help improve interpretability, but may not provide complete understanding.

• Trade-offs may be necessary between interpretability and performance.

Lorem ipsum dolor sit amet

LIMITATIONS

Lorem ipsum dolor sit amet

Difficulty with unstructured data:

• ANNs may struggle to process unstructured data, such as text or images, as they require the data to be transformed into a numerical form.

• In CIM systems, unstructured data may be generated from sources such as text descriptions or images of products or processes.

• Specialized techniques may be needed to extract meaningful features from unstructured data for use in training the ANN.

• Preprocessing and feature engineering can be time-consuming and may require domain expertise.

LIMITATIONS

Lorem ipsum dolor sit amet

limited ability to handle complex relationships:

• ANNs may not be able to capture complex, non-linear relationships in the data, which can limit their performance on certain tasks.

• In CIM systems, complex relationships may be present in the data due to the nature of the manufacturing process or the products being produced.

• Alternative machine learning models, such as decision trees or support vector machines, may be more suitable for handling complex relationships.

• Ensemble methods, which combine the predictions of multiple models, can also be used to capture complex relationships.

LIMITATIONS

Lorem ipsum dolor sit amet

Ethical considerations:

• ANNs can be biased if the data used to train them is biased.

• In CIM systems, the decisions made by the ANN can have significant impacts on people or organizations.

• It is important to ensure that the data used to train the ANN is representative and unbiased, in order to avoid perpetuating existing biases or discrimination.

• It may also be necessary to consider the ethical implications of the decisions made by the ANN and ensure that they align with the values of the organization and stakeholders.

LIMITATIONS

Limited transferability:

• ANNs may not perform well on tasks unrelated to the ones they were trained on.

• Transfer learning can be used to adapt pre-trained models to new tasks.

• Performance may be limited if tasks are significantly different or if there is a large gap in data distribution.

Lorem ipsum dolor sit amet

Other Limitations:

• Long training times

• Limited flexibility

• High hardware requirements

THANK YOU!

R.M.A.W.@GMAIL.COM

Lorem ipsum dolor sit amet

+INFO

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Nunc non pharetra ante.