Types of Machine Learning

Machine learning can be a very broad subject and is typically categorized into 5 different types.

supervised learning

unsupervised learning

semi-supervised learning

self-supervised learning

reinforced learning

Supervised learning to simply put is when we gather a set of features (X) and with theses features produce a label (y) which is the answer or prediction or classification. This is where the machine learns form given the right answers. for example lets say we want to know the price of a car, we first gather the data lets say car model, make, mileage(features / X), etc and we predict the price (label / y). we do this by training the our model with a data set of cars and their prices.

some typical types of supervised leaning algorithms are regression and classification

Unsupervised learning is very similar to supervised learning except we don't provide the model with the labels (y) and only the features(X). We can try to get the Ml algorithm to find patterns in the data. For example lets say i feed my algorithm with data about visitors to my site and then try to predict some patterns in the data for example when users visit, their locations, etc and then try to make some predictions about my visitors based on the features provided.

some typical types of unsupervised learning algorithms are clustering, anomaly detection, dimensionality reduction / feature extraction

Semi-Supervised learning is a mix of both supervised and unsupervised learning, where a small part of the data we train the model with is labeled and the majority of the data is unlabeled. This types of learning algorithm are useful when getting accurate labeled data is difficult or expensive, but unlabeled data is easy to acquire.

An example of this is when we have a large dataset of unlabeled data, but don't have enough resources to manually label all the data, so we label some and use that to train our model

Self-Supervised learning is a type of leaning where the data set is unlabeled and is used to label the data.

For example, lets say you want to have a plant classification model, that takes any img of a plant and predicts its type. you can mask part of the image and then train the model to make differentiations amongst the imgs so it can tell the different types of plants. Then you can tweak its parameters and map to a labeled set of plant imgs so then the model can map the types to generate the labels (classifications)

Reinforced learning is type of learning where the leaning system (ai agent) is aware of its surroundings (environment) and choose to do some actions. Each action can either lead to a positive or negative outcome. It then learns what the best action is for that situation by repeating the scenario and figuring out which yeld to the most positive outcomes

for example self driving cars are an example of reinforced leaning, the car is learning which yelds the most positive outcomes (staying in lane, maintaining speed, not crashing)

neural nets, convoluted neural nets are typically used for reinforced learning,