Summary of Machine Learning

Machine learning is a subset of Artificial intelligence. Machine learning focuses on designing a particular system and allowing them to learn and make predictions based on experience. Before going into the summary of machine learning we obviously need to know about at least one programming language like python/R, data structure, database, math, and statistics.

Types

Machine learning is three types

1. Supervised learning, 2. Unsupervised, 3. Reinforcement learning

Simple process

The simple process of machine learning is

1. Collecting and gathering the data then

2. Training the data then

3. Train the ML algorithm then

4. Model input data, after that

5. ML algorithm, now if at that time new data arrives or collected then add in this part, then

6. Make a prediction.

If anything went wrong then go to the 3rd step and process it again.

Supervised learning

The machine learns by using labelled data. Types of the problem are regression and classification. The algorithms used are LR, Logistic regression, KNN, SVM etc.

Unsupervised learning

The machine is trained on unlabelled data. Guidances are not needed in this case. Types of problems are association and clustering. The algorithm which is used is K-means, C-means etc.

Reinforcement learning

An agent interacts with its environment by producing actions and discovering errors or rewards. NO training supervision is needed. The algorithm is Q-learning, SARSA etc.

Data

Data is a fact, stats or numbers which we collect for further analysis. It is actually two types. One is qualitative and another one is quantitative data. Qualitative is dependent upon characteristics with a deviation of Nominal and Ordinal data. Quantitative data depends upon numbers with the deviation of Discrete and Continuous data. Basic terminologies for the data are population and sampling. We further deviate from the sampling process.

Statistics

The two types of statistics we generally follow. One is Descriptive statistics and another one is inferential statistics.

In descriptive, we measure measures of spread, mean, median, and mode. Information gain & entropy for predicting impurity and information for final outcome. Confusion matrix for describing the performance of classification. And probability full.

In inferential we find the estimation. The methods for estimating the data and population are methods of moments, maximum likelihood, Bayes estimator, confidence interval etc. Hypothesis testing for checking the acceptance of the hypothesis.

From a programming language point of view, a summary of the each libraries and their importance for making machine learning algorithms is important. For understanding the data and problem, choosing the right algorithm is important for machine learning techniques.