What is Machine Learning_ - MMA 869_ Machine Learning & AI

1/4/22, 3:10 AM What is Machine Learning? https://smithlearning.ca/d2l/le/content/371662/viewContent/3210750/View 1/4 What is Machine Learning? Machine Learning (ML) is one of the six AI disciplines, and arguably the most important for businesses today. Traditionally, human programmers explicitly tell computer applications how to perform tasks. For example, programmers writing video game logic (without ML) would encode rules via if- then statements like: if the user ±res rocket launcher, then hide behind the wall. The non-ML approach works very well for many (most?) computer applications. However, consider the situation of the programmer not knowing all the rules beforehand. How can they tell the computer what to do? Find out The discipline of machine learning is broken down into three subtypes: supervised learning, unsupervised learning, and reinforcement learning. Each subtype di²ers in the kind of data they need and the results they can produce. Table 1.2 summarizes the di²erences between the subtypes. Supervised Unsupervised Reinforcement What Predict something in the future Find relationships Learn from mistakes machine learning (noun) We de±ne ML as: Performing a task without explicit instructions from a human, instead relying on data to infer patterns and inference

1/4/22, 3:10 AM What is Machine Learning? https://smithlearning.ca/d2l/le/content/371662/viewContent/3210750/View 2/4 Supervised Unsupervised Reinforcement How Algorithm builds model from past data Algorithms ±nds patterns in data Algorithm takes actions, gets rewards Data Labeled Unlabeled None Tasks/ Algorithms • Classi±cation • Decision Tree, SVM, Naïve Bayes • Regression • Linear, Polynomial, Lasso • Deep Learning • Feed forward, RNN, CNN • Clustering • K-Means, DBSCAN, Hierarchical • Recommenders • Collaborative ±ltering, content- based • Social Network Analysis • Centrality, shortest paths • Association rules • Apriori, Eclat, FP- Growth • Q-learning • SARSA • Deep Q Network Table 1.2. Di±erences between the three subtypes of machine learning An important distinction to make is the di²erence between labeled and unlabeled data. But what is a label? Brie³y, a label is an answer for a supervised algorithm to learn.

1/4/22, 3:10 AM What is Machine Learning? https://smithlearning.ca/d2l/le/content/371662/viewContent/3210750/View 4/4 Figure 1.3. The main terminology in ML. Synonyms are shown in parenthesis, but the bold, black text are the terms used most often Google also provides a clear and comprehensive glossary of ML terms (https://developers.google.com/machine-learning/glossary/) that you can read through if you’re interested in learning more. 