On the 24th of October 2017, I took part in a seminar by Vincenzo Lagani hosted by the R-Ladies Tbilisi group. Vincenzo gave a broad introduction to machine learning with applications in R.
The speaker covered philosophical, theoretical and practical questions about the topic. It looks like I was interested mostly in the first aspect. My findings are the following (it is my interpretation of Vincenzo’s message mixed with my experience, and I can be wrong in my conclusions).
Machine Learning (ML), Data Mining (DM), Deep Learning (DL) and Artificial Intelligence (AI) are, in many cases, the same sort of thing. These labels mark highly overlapping parts of data science. These labels are popular and heavily promoted on the market. But they don’t have standard definitions.
Data Mining is now a fading star in mass media but is the widest term incorporating Machine Learning and friends. Data Mining includes analysis of a decision maker’s problem. In contrast, Machine Learning is a process of choosing a statistical model and its parameters for a given purpose and a data set.
Intuitively Machine Learning differs from Statistical Learning by moving the focus toward the computation aspects of the process. In Statistical Learning, we train our mathematical model to deal with our data set. Theoretically, we can do it even with pen and paper. In Machine Learning, we do the same task but employ computer power.
Deep Learning is a subset of Machine Learning focused on Neural Networks and powered by vast usage of computational power. Fast computers with huge amounts of memory allow Deep Learning to employ thousands of layers and to perform numerous iterations to choose the best starting (hyper) parameters. Internally Neural Networks model consists of sets of transformed linear models. In Deep Learning, we have myriads of them.
Where is Artificial Intelligence? Turn around, it is watching you! Sure, your autonomous car is observing. But the board computer feeds situational data to a statistical model trained under the careful attention of Elon Mask’s employees.
Alexander Matrunich originally wrote this post for Rstat.Consulting.