Introduction to Conceptual Overview of Neural Networks
For Data Science & Machine Learning
To solve real-world problems (like Object Recognition, Audio Analysis, Medial Image Segmentation etc.) using advanced Deep Learning (DL), understanding the fundamentals of Neural Networks (NN) is a necessity.
NNs are a robust approach to approximate real-valued, discrete-valued, and vector-valued target functions. NNs are built using simple units that receive input(s) and provide output(s) with these simple units connected to each other in different capacities.
One simple unit in a NN is called a perceptron. The perceptron combines the weights of the input vector linearly which is passed through an activation function and then compared to a threshold value.
The combined perceptrons in a NN are used to search the best-fit hypothesis space of all possible vectors of the target concept. This search is performed using gradient descent (which is the basis of the backpropagation algorithm).
