If you’re new to Machine Learning and its applications, you’ve probably come across several technical phrases that are tough for a rookie mathematician or scientist to comprehend.
Machine Learning
A branch of computer science and artificial intelligence (AI) concerned with the development of systems that can learn from and make predictions based on data.
Instead of being explicitly programmed to perform a task, machine learning allows computers to behave and make data-driven decisions.
When exposed to new data, machine learning programs learn and improve over time.
In recent years, machine learning has been at the heart of a number of technical breakthroughs. It includes self-driving cars, computer vision, and speech recognition systems.
Supervised Learning
A pre-defined dataset “trains” an algorithm.
When given new data, the algorithm can make reliable conclusions based on its training data.
To train a sentiment analysis classifier, use a training set of human-tagged positive, negative, and neutral tweets.
Unsupervised Learning
When a program has a dataset, it can detect patterns and correlations in that dataset automatically.
Example: Clustering is the method of analyzing a dataset of emails and automatically grouping related emails by topic without any prior knowledge or training.
Classification
Classification is a sub-category of Supervised Learning. It is the process of assigning a label to some form of input.
When making discrete, or “yes or no” predictions, classification systems are common.
Mapping a photograph of a person to a male or female classification, for example.
Decision Trees
Decision tree is a decision-making aid that employs a tree-like graph or model of options and their potential outcomes. A decision tree can also depict an algorithm visually.
A decision tree showing survival of passengers on the Titanic (“sibsp” is the number of spouses or siblings aboard). Source: http://en.wikipedia.org/wiki/Decision_tree_learning
Generative Model
A Generative Model is a model that creates data values when some factors bury in probability and statistics.
In Machine Learning, generative models are used to model data directly or as a step on the way to constructing a conditional probability density function.
In other words, you model p(x,y) to make predictions (which may be converted to p(x|y) using the Bayes rule) and to produce likely (x,y) pairs, which is common in Unsupervised Learning.
Naive Bayes, Latent Dirichlet Allocation, and Gaussian Mixture Model are examples of Generative Models.
Discriminative Model
Conditional models, also famous as discriminative models, are a type of machine learning model that represents the dependency of a variable y on a variable x.
These models are frequently employed in Supervised Learning because they attempt to determine conditional probabilities, i.e. p(y|x).
Logistic Regression, SVMs, and Neural Networks are among examples.
Deep Learning
Deep learning is a group of machine learning methods that frequently use Artificial Neural Networks to construct models, and it was a “hot issue” in recent years.
Because of their capacity to identify the best features and express layers of representation, deep learning approaches have been particularly successful in tackling Image Recognition challenges.
Neural Networks or Artificial Neural Networks
A simple Neural Network. Source: http://en.wikipedia.org/wiki/Artificial_neural_network
Artificial neural networks are a network of interconnected nodes that make up a model, inspired by biological neural networks.
They’re statistical learning models for estimating or approximating functions with a large number of inputs.
When the number of inputs is far too vast for the previously stated machine learning algorithms, neural networks are typically used.
Click here to read more useful and interesting articles.