Editing Neural Network

= Neural Network =

'''Neural Networks''' are a class of algorithms within [[Machine Learning]] and [[Deep Learning]] that are designed to recognize patterns. They are inspired by the structure and function of the biological brain and are used to approximate complex functions by learning from data.

== Overview ==
A neural network consists of interconnected units (called '''neurons''' or '''nodes''') organized in layers. These layers process input data through weighted connections, applying transformations to uncover patterns, make decisions, or generate outputs.

Neural networks are the foundation of many advanced artificial intelligence applications, such as image recognition, natural language processing, and autonomous systems.

== History ==
The concept of artificial neurons dates back to the 1940s with the introduction of the McCulloch–Pitts model. In the 1950s and 1960s, early models like the Perceptron were developed. The field saw renewed interest in the 1980s with the invention of the backpropagation algorithm, and significant breakthroughs occurred in the 2010s with deep learning and GPU computing.

== Structure of a Neural Network ==

=== Layers ===
* '''Input Layer''': Receives raw data.
* '''Hidden Layers''': One or more layers where computations occur; each layer extracts features.
* '''Output Layer''': Produces the final result, such as a classification or prediction.

=== Neurons ===
Each neuron receives input, applies a transformation (using a weight and bias), and passes the result through an '''activation function'''.

=== Activation Functions ===
* Sigmoid: `f(x) = 1 / (1 + e^{-x})`
* ReLU (Rectified Linear Unit): `f(x) = max(0, x)`
* Tanh: `f(x) = (e^x - e^{-x}) / (e^x + e^{-x})`
These functions introduce non-linearity, allowing the network to model complex patterns.

== Training Neural Networks ==

=== Forward Propagation ===
Data passes through the layers to compute an output.

=== Loss Function ===
Measures the error between the network’s prediction and the actual output (e.g., Mean Squared Error, Cross-Entropy).

=== Backpropagation ===
A method to update the weights using gradient descent, minimizing the loss function by propagating errors backward through the network.

== Types of Neural Networks ==

=== Feedforward Neural Networks (FNN) ===
The simplest type, where information moves in one direction from input to output.

=== Convolutional Neural Networks (CNN) ===
Used primarily for image and video recognition. They apply filters to detect spatial hierarchies in visual data.

* See also: [[Deep Learning]]

=== Recurrent Neural Networks (RNN) ===
Designed to handle sequential data by maintaining internal memory. Useful in time series and language processing tasks.

=== Long Short-Term Memory (LSTM) ===
A type of RNN capable of learning long-term dependencies. Common in [[Natural Language Processing]].

=== Generative Adversarial Networks (GANs) ===
Consist of two networks (generator and discriminator) trained in opposition to generate realistic data.

=== Transformers ===
State-of-the-art architecture for sequence modeling, using self-attention rather than recurrence.

== Applications ==

* Handwriting recognition
* Speech-to-text systems
* Image classification
* Language translation
* Fraud detection
* Autonomous vehicles

== Advantages ==

* Capable of modeling complex, non-linear relationships
* Learns features automatically from raw data
* High performance in vision, speech, and text tasks

== Limitations ==

* Requires large amounts of data and computational power
* Difficult to interpret (black box)
* Prone to overfitting if not properly regularized

== See Also ==
* [[Artificial Intelligence]]
* [[Machine Learning]]
* [[Deep Learning]]
* [[Data Science]]
* [[Natural Language Processing]]
* [[Backpropagation]]

== References ==
<references />

[[Category:Neural Networks]]
[[Category:Machine Learning]]
[[Category:Artificial Intelligence]]
[[Category:Deep Learning]]