What is machine learning
What is Machine Learning and Artificial Intelligence? What are the use cases of Artificial Intelligence and Machine Learning?
You may have heard a lot recently about Machine Learning, one of the most prominent subfields of AI, but it is not a new concept that emerged overnight. It became popular back in the 1990s, with the notable win of the chess-playing computer Deep Blue over world champion Garry Kasparov. However, it was the last decade that brought Machine Learning into the spotlight, with applications in various fields, ranging from health care, education, and especially business. But what does Machine Learning actually do and why is it so powerful? This article provides answers to all your questions and curiosities about the fascinating field of Machine Learning.
What is Machine Learning?
Arthur Samuel, a pioneer in the field of AI, defined “Machine Learning” as the “field of study that gives computers the capability to learn without being explicitly programmed”.
Using Machine Learning, computers are capable to learn and automate by themselves based on their experiences, without requiring programming from humans. Machines can be trained with machine learning models, built using quality data and various algorithms, depending on the type of data and the task to be automated.
The way in which machines learn is similar to the human learning. Consider the example of studying for an exam: students “feed” their brain with quality information (from books and courses), while also assimilating the desired output (using already-solved questions to understand the approach they must adopt). Over time and with the right amount of practice, their performance improves substantially. Similarly, machines are trained with both inputs and outputs, until the model is capable to use only the input to produce an accurate output that has not been fed during training.
Is Machine Learning the same as Artificial intelligence?
Although the world is buzzing around the groundbreaking development of Artificial Intelligence and Machine Learning, there is a clear distinction between the two terms.
AI is a broad field comprising all the instances when intelligence is added to the machine’s capabilities, making them “think” in a similar way as humans. AI is responsible for the revolutionary technologies, including self-driving cars, gaming and personal assistants. Machine Learning is only a subfield of AI and refers to the study of machines that learn, adapt and improve automatically from data and previous experiences. Due to this capability, Machine Learning is especially performant in recognizing patterns and making predictions. To summarize, Machine Learning is Artificial Intelligence, but not all AI is Machine Learning.
How is Machine Learning different from Traditional Programming?
To obtain the output in traditional programming, the machine is fed with data (input) and with the program that contains the logic through which the input must be analyzed. But using machine learning, the input and output are used to train the machine, which creates its own logic that can render output using only the input.
How can machine learning techniques be classified?
There are two broad classifications to which machine learning problems can be assigned: supervised and unsupervised learning.
Requires the presence of a supervisor as a teacher to train the machine using data already tagged with the correct answer. Afterwards, when the machine is provided with new input data, the supervised learning algorithm analyses the data used for training and selects the correct outcome from it. Thus, the machine learns the characteristics of the training data and then apply the knowledge to new data. Supervised learning uses two types of algorithms:
- Classification: when the output variable is a category (e.g. “disease” and “no disease”).
- Regression: when the output variable is a real value (e.g. “dollars” or “weight”).
Refers to the training of machines using non-labeled information and allowing the algorithm to act on that information without guidance. The task of the machine is to identify similarities, patterns and differences that can be used to categorize the unsorted information, without any prior training of data. Unlike supervised learning, there is no teacher to provide training, therefore the machine must find by itself the hidden structure in the unlabeled data. Unsupervised learning employs two categories of algorithms:
- Clustering : discover the inherent groupings in the data (e.g. grouping customers by purchasing behavior).
- Association : discover rules that describe large portions of your data (e.g. people that buy X also tend to buy Y).
What is Machine Learning used for?
Machine Learning is widely used in the business context, being in charge of processing, sorting, and finding patterns in large amounts of data, at a large scale that would be unfeasible for humans. All the prestigious companies in the technological industry (such as Microsoft, Google, IBM) developed and offer tools based on machine learning.
Some subsets of Machine Learning
Some of the most important Machine Learning developments are explained in the upcoming subsections:
Natural Language Processing (NLP)
Natural Language Processing is a component of AI that represents the ability of a computer program to process, analyze and understand human language as it is spoken. NLP studies the interactions between machines and humans, via linguistics. Some of the most common use cases of NLP include:
- Answering various questions asked by humans (e.g. Siri, Alexa, and Cortana);
- Mapping image to text (create captions from an input image);
- Sentiment analysis (label an attitude as positive, negative, or neutral);
- Speech recognition;
- Translate text into different languages.
Computer Vision is a subfield of AI which deals with how machines interpret the real world. Frequent uses of computer vision include Facial, Pattern and Character Recognition techniques. Computer vision is also increasingly used to automatically analyse images or documents.
Some use cases of Machine Learning
Increasingly becoming a “must-have” in online retail websites such as Amazon.com, this Machine Learning algorithm learns from the user’s personal preferences and previous browsing and purchase activity to make recommendations that would tempt them to make another purchase. The use of Recommender Systems has expanded outside retail, with the feature being implemented by media providers and social networks, such as Netflix, Youtube and Instagram.
Pricing and Stock Market trends
In order to better assess the market, Machine Learning systems such as regression techniques come in handy, especially for price prediction and stock market trends analysis. This is what algorithmic trading is all about.
One new field of application of AI is in the parsing of invoices, such as what Fyn is doing. By using Machine Learning capabilities to the initial OCR layer, Fyn acquires the capability not only to extract data from specific fields in your financial documents, but also to learn from experience and improve the algorithm in order to provide more accurate data extraction. Fyn relies on a blend of AI techniques that outperform existing market solutions.