Evolution of Machine Learning

Because of new computing technologies, machine learning today is not like machine learning of the past. During its nascent stage, machine learning was concerned more with mathematical theory, and intelligent, strong, efficient algorithms. This makes sense given the computational constraints of the time. Recently, huge computational power, and an exponential increase in space and time has changed the face of the field. Machine learning is still interested in intelligent, powerful algorithms, but it’s less concerned with efficiency. Huge algorithms that can learn from huge sets of examples have taken the place of quick learning, small algorithms. These large algorithms tend to be capable of performing more operations, and we’re seeing more and more single methods that can do everything that’s talked about in ML. Machine learning is not a new science, but it’s only beginning to ripen.

Here are a few widely publicized examples of machine learning applications you may be familiar with:

• Self-driving cars: Regression with reinforcement learning
• Recommender systems, like those found in Amazon and Netflix: Clustering methods, predictive classifiers, and a whole lot of semi-supervised
• Twitter scrappers: Natural language processing based in classification and regression models, with semi-supervised learning.
• Fraud detection: Classifiers and regressors with supervised learning.