What is Machine Learning, Simply?
Machine learning, or rather the idea machines can learn to ‘do’ without an explicit set of instructions (programming), has been the basis of many movies where humans end up getting the short end of the deal. But is machine learning truly that dire?
Unlikely. Machine learning, which is a subcategory of artificial intelligence, is simply a way for machines to imitate intelligent human behavior. It’s a type of data analysis that allows programs to learn via experience in order to complete complex tasks, much like humans problem-solve. This type of learning typically breaks down into two specific types: deep learning and reinforcement learning. But what’s the difference?
Deep learning is essentially what you see in any young child as they start to understand that while chickens are birds, not all large birds are chickens. It is based upon the ability to classify both the common features (in this case: feathers, beaks, wings, etc) as well as the uncommon features that separate each grouping from each other (sound, size, feather pattern, beak length). This kind of hierarchical feature learning stacks multiple layers of learning nodes as observed data from one layer produces new outputs that are then fed to a higher level.
In deep learning, the machine begins with raw data that must then be sorted into relevant and irrelevant subsets. The machine, exposed to more data, improves over time. This is similiar to how a baby learns.
Meanwhile, reinforcement learning relies more on trying out slight variations of a problem. As results occur (favorable and unfavorable) data sets change until the best outcome emerges. This is reminiscent of “The Good Place” as Michael tries to create a better version of his neighborhood.
Reinforcement learning uses a closed-loop algorithm where each action receives feedback in a trial-in-error process until the best action is determined.
Machine Learning vs. Artificial Intelligence
While the two things are connected, machine learning is not the same thing as artificial intelligence, or AI. Artificial intelligence is the science of creating computers that behave with the same capabilities as the human brain. This end target requires many technological advances, including machine learning, which uses algorithms that help computer programs automatically improve using experience (as noted above.) But ML is only one part of AI.
How does Machine Learning Work
Machine learning uses algorithms to process datasets in order to learn and improve accuracy. Supervised learning models offer computers labelled training data. Unsupervised learning models use unlabeled data. There are also models of training that fall between the two called semi-supervised learning.
Supervised learning maps an input to an output based on example input-output pairs. A classic example of supervised learning is predicting house prices. Data in the form of square footage, features, number of rooms, location, etc are fed into the model. Data training can lead to accurate home price predictions from data inputs.
Another example is weather prediction. By using historical temperature, precipitation, wind, and humidity data, better predictions of future weather can be made.
Semi-supervised learning takes the same approach as supervised learning, but with less data labelling. For example, in semi-supervised learning, only 30% of 10,000 cat and dog images fed into a computer will include a “cat” or “dog” label.
Unsupervised learning looks for previously unknown or undetected patterns to find clusters of data with commonalities. This kind of pattern recognition is a cliche in some bad tv shows, where it seems like certain characters have nearly superhuman abilities to connect disparate data. Think of Sherlock, House, The Blacklist, or any cop show.
In machine learning, unsupervised learning reduces the complexity of the problem by reducing the number of random variables needed to solve the problem. Algorithms cluster data in order to further reduce complexity.
FAQs about Machine Learning
- Q: What’s the best programming language for machine learning?
- Q: Why call it machine learning?
- A: Because computers learn how to perform tasks without specific programming for those tasks. Instead, they’re using training data and analytics to learn how to respond. ML is similar to but not exactly like statistical learning, which tends to rely on rule-based programming, a smaller dataset, and is assumption-dependent.
- Q: What are some machine learning uses/examples?
- A: Image and speech recognition. Medical diagnosis. Predictions within complicated systems(like worldwide pandemics). Turning unstructured information into structured data.
- Q: Why is machine learning important/what are the benefits?
- A: As a subset of AI (artificial intelligence), it allows us to use more data in less time. Applications use whatever data we provide, and will identify trends and patterns humans might otherwise miss.
- Q: What are the objectives of machine learning?
- A: The primary objective of machine learning is pattern recognition within data, leading to predictive problem-solving based upon that knowledge.
- Q: Does Google use machine learning?
- A: Yes, within Google Translate, in its Google Search feature, as well as in Google Photos and in Gmail.
- Q: What jobs use machine learning?
- A: Human-Centered ML Designer. Computational Linguist. Data Scientist. ML Engineer. Software Developer. Software Engineer. Business Intelligence Developer.
- Q: Is there a demand for workers who know machine learning?
- A: The short answer is, yes. Many jobs are moving out of the IT department and into other departments. According to Gartner, while the number of AI/machine learning jobs posted by IT departments between 2015 and 2019 tripled, that’s significantly less than the jobs posted by other types of business areas. (1)
Other Definitions Used In Machine Learning
- Algorithm: a method that will allows system learning and improvement from experience alone, without explicit programming.
- Bias: Also known as AI bias or algorithm bias, this occurs when observed results have been systematically prejudiced due to false or imprecise assumptions.
- Kernel: Also referred to as a kernel function or a kernel trick. A kernel function acts as a bridge from linearity to non-linearity in algorithms. There are a number of different kernel functions including linear, Gaussian, polynomial, exponential, sigmoid, and laplacian. More reading about kernels, here: https://arxiv.org/pdf/math/0701907.pdf
- Regression: Linear regression is a type of machine learning algorithm that is built upon a supervised learning model. Regression predicts a dependent variable value (denoted by y) based upon an independent variable (denoted by x.)
- SVM: SVM stands for Support Vector Machine, and is also sometimes called support-vector networks. SVMs are supervised learning algorithms. They are most often used for regression and for classification.
Need machine vision options for your factory? Take a look at our inventory of Cognex Insight Cameras.