Computer methodology - Artificial intelligence

Diagram of a simple feed-forward artificial neural network, with one “hidden layer,” also known as a “perceptron.” Image: Wikipedia

Deep learning simply means “stacked neural networks” which are networks composed of several layers (Pathmind, 2021). Based on Rosenblatt’s perceptron, the simplest version of an artificial neural network has three layers of neurons. The first is the input layer which takes input values ,which in other words is data. This first layer of neurons is connected to the “hidden” layer. The outputs of these “hidden” neurons are then connected to the final output layer. This final layer is what gives you the answer to what the network has been trained to do (Medium, 2021). 

The simplest neural networks consist of input layer, hidden layer and output layer, which are attached to each other with predictors, these predictors come attached with coefficients called "weights" , this is a non-linear network known as multilayer feed-foward network (Hyndman, R.J. and Athanasopoulos, G, 2018). Neurons are found in the hidden layer of neural networks and in supervised learning neurons are provided with labelled data which allows neurons to build knowledge from data with the correct answer in advance the networks then gradually learn  to find the right answer on their own, increasing the accuracy of their predictions (KDnuggets, 2021). One methodology for deep learning learning is the previously mentioned  feed forward network, this model allows the signals move in one direction, towards the output layer, they are widely used in pattern recognition(Upgrad, 2021).

According to a study published by Mehrotra et al.(1997) as cited in Stanford university articles (2021) Feed-forward networks have the following characteristics:

1. Perceptrons are arranged in layers, the first layer consists of inputs and the last layer is outputs. The middle layers have no connection with the external world, thus they are called hidden layers.

2. The information is constantly "fed forward" from one layer into the other because each perceptron in one layer is connected to every perceptron on the next layer. 

3. Perceptrons in the same layer have no connection.

Reference:

 kdnuggets, 2021. How do Neural Networks Learn? - KDnuggets. [online] KDnuggets. Available at: <https://www.kdnuggets.com/2015/12/how-do-neural-networks-learn.html#:~:text=Neural%20networks%20generally%20perform%20supervised,the%20accuracy%20of%20their%20predictions.> [Accessed 20 May 2021].

Pathmind. 2021. A Beginner's Guide to Neural Networks and Deep Learning. [online] Available at: <https://wiki.pathmind.com/neural-network> [Accessed 20 May 2021].

Medium, 2021. How Do Neural Network Systems Work?. [online] Medium. Available at: <https://medium.com/chmcore/how-do-neural-network-systems-work-dbe1bc0c4226> [Accessed 20 May 2021].

Hyndman, R.J., & Athanasopoulos, G. (2018) Forecasting: principles and practice, 2nd edition, OTexts: Melbourne, Australia. OTexts.com/fpp2. Accessed on 20/05/2021.

Upgrad, 2021. Neural Network: Architecture, Components & Top Algorithms | upGrad blog. [online] upGrad blog. Available at: <https://www.upgrad.com/blog/neural-network-architecture-components-algorithms/> [Accessed 20 May 2021].

Stanford, 2021. Neural Networks - Sources. [online] Cs.stanford.edu. Available at: <https://cs.stanford.edu/people/eroberts/courses/soco/projects/neural-networks/Sources/index.html> [Accessed 20 May 2021].