These 10 keywords make you talk like a machine learning expert!
Ever heard a conversation between two machine learning engineers? Did you get any of what they were saying? Of course you didn’t! Would you like to? Here are 10 keywords that probably won’t make you an expert in machine learning, but will at least light a tiny bulb in your head when they mention them so you that you can at least nod in the right places.
1. Machine Learning
Machine learning is a type of artificial intelligence (we’ll get to that as well later in this article) that allows computers to learn something by being exposed to new data, rather than by being explicitly programmed for the task.
2. Artificial Intelligence
Artificial intelligence or AI is a sub-field of computer science whose main goal is to create intelligent machines (ones that would not need to be programmed for a specific task like what we have today). In simple terms, AI aims to make machines as closely resembling humans in their behavior as possible.
3. Deep Learning
Deep learning is a branch of machine learning inspired by the function of the brain called “artificial neural networks” (ANN). Basically, since machine learning has strayed from its original objective of creating AI, deep learning has the task of bringing ML back to its roots, so to speak.
The goal of classification is to build models that separate data into different classes. The models are built by first pre-labeling classes for a set of training data and introducing that data into models, thus allowing the algorithm to learn from the training data.
Sometimes, data does not have a class attribute or at least doesn’t include pre-labeled classes. These types of data are in that case grouped together via “maximizing the intraclass similarity and minimizing interclass similarity” (Han, Kamber and Pei). In other, simpler words, data gets clustered (grouped) by similarity.
6. Support Vector Machines
Support vector machines (SVM) take training data and transform it into a higher dimension, which is in turn inspected for the optimal separation boundary (or boundaries) between individual classes. These boundaries are called “hyperplanes” and the idea behind SVM is that, with enough dimensions, it will always be possible to find a hyperplane separating two distinct classes.
7. Bayesian Probability
Bayesian probability (named after Thomas Bayes) refers to use probability in an attempt to predict how likely certain events will happen in the future. It is commonly expressed as a percentage, unlike traditional probability, which uses frequency.
8. Data Mining
Data mining involves examining large sets of data and locating patterns, correlations and anomalies within them and predicting outcomes based on that data. It can be used to predict future market trends and allow businesses to make better decisions.
9. Reinforcement Learning
According to Christopher Bishop, Laboratory Director at Microsoft Research in Cambridge, reinforcement learning is “concerned with the problem of finding suitable actions to take in a given situation in order to maximize a reward”. Reinforcement algorithms do not get specific goals, but have to learn them by trial and error.
10. Neural Networks
Neural networks represent sets of algorithms based on the human brain. Their task is to interpret sensory data, labeling and or clustering raw input they get and recognize patterns. These patterns are numerical and all real-world data (text, sound, images, etc.) must be translated in vectors.
Did you get any of that? Good, now you can talk like a machine learning expert.