Neural computing
Presented By:
Sandesh Thapa Magar
M.E Computer
1
Neural Computing
Neural computing mimics biological neurons to create machine-learning models capable of
learning from experience, improving over time, and solving complex problems.
2
Structure of an Artificial Neural Network (ANN)
A typical ANN consists of:
[Link] Layer – Receives raw data (e.g., an image, text, or numerical data).
2. Hidden Layers – Process and extract meaningful features from data.
[Link] Layer – Produces the final result (e.g., classifying an image as "cat" or"dog")
3
Neural Computing
4
Neural network category
[Link] neural networks usually have only one hidden layer
[Link] neural networks have multiple hidden layers
5
Type of Neural network
[Link] Forward Neural Networks (FNNs)
[Link] Neural Networks (CNNs)
[Link] Neural Networks (RNNs)
[Link] Short-Term Memory (LSTM) Networks
6
[Link] Forward Neural Networks (FNNs)
A Feedforward Neural Network (FNN) is one of the simplest types of artificial neural networks
where information flows in one direction from input to output, without any feedback loops.
7
[Link] Neural Networks (CNNs)
Convolutional Neural Networks (CNNs) are a specialized type of deep neural network designed for
image processing and pattern recognition.
They are particularly powerful for tasks involving visual data, such as image classification, object
detection, and segmentation.
8
[Link] Neural Networks (RNNs)
An RNN is a type of neural network designed for processing sequential data (e.g., text, speech).
Unlike traditional neural networks, RNNs have loops that allow information to persist across time
steps, meaning they can "remember" previous inputs in the sequence.
9
[Link] Short-Term Memory (LSTM) Networks
● LSTM is a type of Recurrent Neural Network (RNN) designed to solve the problem of
learning long-term dependencies in sequential data.
● It improves on traditional RNNs by preventing the vanishing gradient problem, which
makes it harder for standard RNNs to remember information over long sequences .
10
Learning of a Neural Network
Neural networks learn by adjusting their internal parameters based on input data. The
learning process can be classified into three main types:
1. Supervised Learning
In supervised learning, the neural network learns from labeled data.
● Input: Features (e.g., images of cats and dogs).
● Output: Known labels (e.g., "cat" or "dog").
● Goal: Minimize the difference between predicted and actual output using a loss
function.
Examples of Supervised Learning:
Image classification (CNNs for recognizing objects).
Spam email detection.
Medical diagnosis (predicting diseases based on symptoms).
11
Learning of a Neural Network
2. Unsupervised Learning
In unsupervised learning, the network learns patterns without labeled data.
● Input: Unlabeled data (e.g., customer purchase behavior).
● Output: Hidden patterns or clusters (e.g., grouping similar customers).
● Goal: Identify structures in data without explicit labels.
Examples of Unsupervised Learning:
Customer segmentation (grouping shoppers based on behavior).
Anomaly detection (fraud detection in banking).
12
Learning of a Neural Network
[Link] Learning (RL)
In reinforcement learning, the network learns through trial and error by interacting with an
environment.
● Agent: The model that makes decisions.
● Environment: The world where the agent operates.
● Rewards: Feedback to encourage good decisions and discourage bad ones.
Examples of Reinforcement Learning:
Self-driving cars (learning how to navigate roads).
Game playing(like Chess).
Robotics (autonomous robots learning tasks).
13
Applications of Neural Computing
[Link]-Driving Cars – Neural networks help recognize objects and make driving decisions.
[Link] Processing – Facial recognition, medical imaging.
[Link] Language Processing (NLP) – Chatbots, speech-to-text.
4. Finance & Security – Fraud detection, stock market prediction.
[Link] – Disease prediction, drug discovery.
14
Thank You
15