Machine learning is classified into two types.
1. Supervised learning
2. Unsupervised learning
Most of the persons think that machine learning has one more field that is reinforcement learning. yes there are two more terms that is reinforcement learning and recommendation but these two terms are totally belong to deep learning and neural networks. As I told in the
Post 1 that deep learning is the subset of machine learning that is why we consider three types.
1. Supervised learning
2. Unsupervised learning
3. Reinforcement learning
1. Supervised Learning
In this field of machine learning we have enough information about the data sets or labels and we take these data sets as an input into an algorithm, test and train data set and predict the outcomes.
labelled dataset means we have enough information about dataset and it has both input and output parameters.
"when the model is getting trained over the labelled dataset is know as supervised learning."
Regression and classification are the two approaches that are used in supervised learning.
In Regression, we give the datasets to the algorithm and predict the result and our result contains continuous values.
for
example- any of your friend wants to sell his house so in what amount he should sell his house.
this type of problems can be solved using regression.
On right hand side, there is a figure in which a straight line and a quadratic function is set over data points.
Classification is similar to regression but in the place of continuous value , we get discrete values as an output.
example- we want to predict or check whether any person has breast cancer or not. for this we have tumor size as our dataset. and we make two assumptions for this case that is 0 for benign( normal) case and 1 for malignant case .These types of problem are stated as classification problem.
generally we use test and train method to solve supervised learning problem.
In above mentioned two examples there are limited number of attributes . In case that there are infinite number of attributes so we use SVM that is support vector machine.
2. Unsupervised learning
In this type, data doesn't have any labels or really no labels, so we are given dataset and we're not told what to do with it and we're not told what each data point is.
"Unsupervised learning allows us to approach problems with little or no idea what our results should look like. We can derive structure from data where we don't necessarily known the effect of variables."
we can derive this structure by clustering the data based on relationships among the variables in the data.
Clustering is the example of unsupervised learning.
for example- Take a collection of 1000000 different genes, and find a way automatically group these genes into groups that are somehow similar or related by different variables.
In clustering, unlabeled data provided but in classification, labeled data provided that is why these two put in unsupervised and supervised learning respectively. In both cases, we make group or cluster with coherent groups of data points that's why we can use k mean cluster algorithm in both cases.
3. Reinforcement learning
learning in neural networks means continuously updating the weights between the neurons of any specific network.
Reinforcement learning is quite similar to supervised learning as we make decisions same as we make prediction in supervised learning.
"Train neural networks model by the actions that agent perform and update set of actions in dataset."
In reinforcement learning, policy networks are present that can be implemented by policy gradients.
Agent calculate the bunch of experience.
After performing every action agent get a reward , if the value of reward is positive , we add that experience in dataset. In case we receive a neg gradient means that action is not very good for our final outcome.
Actions having negative record slowly filtered out from the program so probability of getting failure is very less so only positive actions are there. That's how program run by itself.
A term that generally used in reinforcement learning is CAP (credit assignment problem). It is the process of identifying among the set of actions chosen in an event that ones which are responsible for the final outcome.And moreover it is the best and worst set of decisions chosen over an event so that the best decisions are reinforced and worst penalized.
Now I hope that you have clear view about the types of machine learning.
If have any doubt, just drop a comment.
Connect with our Data Science community from the below mentioned link.
Thank you.
Comments
Post a Comment