that the algorithm learns accurately without deviating from the correct path, which is the essence of supervised learning. Unsupervised learning Unsupervised learning algorithms identify patterns, similarities, and differences in large datasets that might not be visible to humans. Mathematicians discovered long ago how to determine correlation between a single dependent and independent variable, and even how to use advanced statistical techniques to find a relationship between two or three independent and one or two dependent variables. But sometimes the relationship between the variables is obfuscated by hidden variables. For those cases, machine learning takes over in order to determine hidden relationships. By feeding the raw input data into the model and letting the algorithms experiment, it is possible to determine relationships,
Machine learning basics
Types of learning
When laymen hear the term artificial intelligence, their thoughts often turn to the intelligent, personified machines and humanoid robots often seen in movies and television – machines capable of completing any task, who inevitably turn on their creators in a bid to take over the world. Reality is quite different – machines are only able to learn one task at a time, and once their training stops, their ability to evolve stops with it. Machine learning ‘intelligence’ is restricted to the completion of a single task and is frozen in time. For example, a camera trained to detect vehicle license plate numbers, can only ever detect license plate numbers. It cannot ‘evolve’ to edit the grammar in written documents. What is machine learning? Machine learning is one type of artificial intelligence that uses advanced algorithms and a trove of data to teach computers to make predictions. Unlike the omnipotent robots of science fiction, machine learning models can only perform specific, isolated tasks, such as image recognition and classification, language translation, or trend detection. There are three types of machine learning: supervised learning, unsupervised learning, and reinforcement learning.
Supervised learning Supervised learning involves pairing known inputs with known outputs. For example, how do computers learn to recognize handwritten digits when each person’s handwriting is unique? The MNIST database (http://yann. lecun.com/exdb/mnist/) contains 70,000 examples of handwritten digits (0-9) collected from the National Institute of Science and Technology Standard Reference Data Special Database 1 and 3. Researchers collected hundreds
of thousands of handwritten characters and meticulously
All about AI/machine learning
labeled each one with the correct digit. This dataset is used to train computer models to automatically recognize digits. By using this dataset, computer scientists can train and test their algorithms with known correct answers. This approach ensures
Ants are fascinating creatures. When they leave their nest in search of food, they initially wander randomly, leaving pheromones along their path. Once an ant finds food, it returns to the nest, reinforcing the trail with more pheromones. Other ants follow this strengthened trail and continue to add pheromones, making it even more prominent. Occasionally, ants wander off the trail and if they discover a shorter path to the food, the trail will gradually shift to follow this new route. Over time, this process transforms a weak and meandering trail into a streamlined
ant superhighway.
Contributed By M ark Hughes
The process of how ants explore their environment is similar to how brains learn, and machine learning algorithms operate: They start by exploring many possibilities; they identify successful outcomes; and then they optimize and reinforce pathways to make the connections more efficient over time. This article covers how machine learning works, its relation to the ant analogy, and encourages you to consider using machine learning in your next project!
Figure 1: The image illustrates unsupervised learning. In unsupervised learning, a large assortment of data is fed into an algorithm whose job it is to find patterns, coincidences, and anomalies. The algorithm sorts the data in a manner that suits it.
Outputs
Input Raw Data
Interpretation
Algorithms
Processing
we get technical
60
61
Powered by FlippingBook