Part 1 — Introduction to Machine Learning
In Machine-learning models are computer algorithms that use data to make estimations (educated guesses) or decisions, a machine-learning algorithm uses data to get better at a specific task. Machine learning models(function/ programs) are shaped by data as opposed to traditional /old model which are programmed by expert A model as a function that accepts data as an input and produces an output. The human designer doesn’t select parameter values, instead they are set to an initial guess, then adjusted during an automated learning process called training. Traditional programming we get the following form and expert who have ben in the field to estimate the
Traditional way
Experience + Domain knowledge = output function ( f(x) =x2 + 3 )
Data Science way
Data + Domain Knowledge +Algorithm = output function ( f(x) =x2 + 3 )
The goal of training is to improve a model so that it can make high-quality estimations or predictions.
Models are trained using data plus two pieces of code, the objective function and the optimizer.
The objective is what we want the model to be able to do; so that a computer can understand our objective, we need to provide our goal as code snippet called an objective function (also known as cost function)
The Objective Function
Objective functions judge whether the model is doing a good job (estimating boot size well) or bad job (estimating boot size badly).
The Data
Data refers to the information that we provide to the model (also known as inputs).
The Optimizer
During training, the model makes a prediction, and the objective function calculates how well it performed. The optimizer is code that then changes the model’s parameters so the model will do a better job next time
Training a model is the process of improving how well a model works, we train once and use multiple time can take a minute or daysAfter training we no longer need the original data, the objective function, or the model updater. When you one use a model you provide inputs and receiving an estimation or prediction, it usually takes a few seconds Taken together, this means that to use a model, we only ever need features, while during training we usually need both features and labels.
