Introduction to Machine Learning

History Of Machine Learning

The first attempt to automate data processing was made by Pascal, who created a mechanical adder that used gears and wheels. A machine he devised could divide and multiply directly by adding and subtracting two numbers. On the calculator, metal wheels showed the numbers from 0 to 9 all-around their circumference. Joseph Marie Jacquard invented weaving looms that used metal cards with holes punched into them to position threads. Jacquard's loom, equipped with an interchangeable punch card system, enabled automatic reproduction of any pattern. These punched cards were used by both Charles Babbage, an inventor of the analytical engine, to feed data into his machine, and by Herman Hollerith, an American statistician to feed information into his census machine. By using Boolean operators (AND, OR, NOR), the binary values of yes and no would be represented as true, false, and binary values as 1 or 0. Herman Hollerith developed a system of mechanical calculations and punch cards to calculate statistics gathered from millions of people. An electromechanical device known as a tabulating machine was used to combine information stored on punched cards. English mathematician Alan Turing developed the Turing Test to determine whether machines are intelligent in the 1940s and 1950s (Fradkov, 2020; Kononenko, 2001; Plasek, 2016). The Fig.1 depicts the history of machine learning

Figure 1.

History of Machine learning

In 1952, Arthur Samuel created a checkers game to teach computers to play. A supervised training model was used to study which moves to lead to winning strategies, and the software was updated to implement those. A perceptron is a type of neural network designed by Frank Rosenblatt. Neurons in the brain are connected in a network, and a neural network is like the brain. As part of the larger program, decisions are made that help solve complex problems. Students at Stanford University invented the Stanford Cart, which is capable of navigating obstacles by itself in a room. Computer software enabled the Cart to navigate crowded spaces on its own, entirely relying on images broadcast via an onboard TV system for all its knowledge. A journal article by Gerald Dejong in 1981 introduced explanation-based learning (EBL) where training examples are used as evidence of prior knowledge in EBL. Upon receiving instruction, a computer program analyzes the training data to form a general rule based on the information it has been told to discard. As early as the 1990s, machine learning was being applied to data mining, adaptive software, web applications, and language learning. Researchers start creating computer programs that analyze huge amounts of data, then draw conclusions-or “learn”-from them. Advances in technology have made it possible for programs to learn on their own as data is introduced without the need for human intervention. Since the turn of the millennium, adaptive programming has become increasingly popular. The programs recognize patterns and learn from experience when they receive feedback from the world. Adaptive programming uses deep learning algorithms to identify objects in images and videos (Andre Esteva, 2017)