Issue link: https://resources.mouser.com/i/1481902
Intelligence at the Edge 6 AI Domains DATA SCIENCE Data science is the study of data for actionable, forward- looking insights. Data scientists collect and analyze data to extract useful patterns and relationships. Then they build and test predictive models that categorize such information to manage risk, boost product efficiency, and maximize business productivity. ARTIFICIAL INTELLIGENCE (AI) AI is a general term for the development of "smart" computing systems that can perform tasks that normally require human intelligence such as speech, vision, recognition, and decision making. It encompasses all applied fields of ML. MACHINE LEARNING (ML) ML is a AI subfield that empowers machines to learn from examples. AI algorithms are trained to extract relevant features within example data so that machines can then make insightful predictions and decisions about new data on their own. DEEP LEARNING OR NEURAL NETWORKS (NN) Deep Learning is a subset of ML algorithms based on layered and interconnected artificial neurons, inspired by the structure and cognitive functions of the human brain. These "neural networks" (NNs) execute a series of calculations on acquired data to perform tasks more quickly, accurately, and reliably than humans—such as learning how to drive a car or identify objects in a picture. ML Goals ML encompasses many different types of algorithms and training methods that employ statistics, probability, calculus, and linear algebra to analyze and visualize data in various ways. An ML algorithm extracts relevant features from training data to produce a predictive model for unseen data. The algorithm only becomes intelligent after being trained on an organization's data. For instance, no amount of algorithm sophistication can predict machine failures without example failures to learn from. ML algorithms start learning in a naïve state without any context. For example, a stopped electric motor may be perfectly normal in some cases but abnormal in other cases. This simplistic example illustrates the importance of context, and the need for sufficiently varied training data so that context is learned by the algorithm.