Machine learning (Artificial intelligence)

Machine learning (Artificial intelligence) 

Small explanation 

A programming language is a language used to give instructions to a computer. These set of instructions is called code. A code takes some input, performs the set of instructions defined in the code to get an output. However in real world processes, the input is given, the output obtained is observed, but the code/logic is unknown. Machine Learning is the science of deriving the logic from the data of input and output observations. To derive the logic, certain mechanisms are used which are called machine learning algorithms.

 

Each machine learning algorithm has 3 components- a set of unknown parameters, a loss and an optimizer. Let's pick them up one by one. The machine learning algorithm specifies a well defined set of operations on the input variables and the unknown parameters to get an output prediction. This output prediction is different from the actual output. This difference is called loss. Our aim is to decrease the loss by changing the parameters. The optimizer calculates the impact of changing the parameters ondecreasing the loss. Accordingly each parameter is tweaked by optimizer to decrease the loss. This process is repeated with more data until the loss is sufficiently decreased. This means that the output prediction and actual output are very close now (as the loss is decreased). This means that the machine learning algorithm can now generate accurate output predictions from input data and hence the code/logic has been derived.