What is machine learning?

Machine learning (ML) is a form of artificial intelligence (AI) that enables software applications to improve their accuracy at predicting outcomes without being explicitly programmed to do so. This is achieved by using historical data as input to predict new output values.

Recommendation engines are a common application of machine learning. Other popular uses include fraud detection, spam filtering, malware threat detection, business process automation (BPA), and predictive maintenance.

Why is machine learning important?

Machine learning is valuable because it allows businesses to understand trends in customer behavior and business operational patterns, and to support the development of new products. Many leading companies, such as Facebook, Google, and Uber, rely on machine learning as a key part of their operations. As a result, machine learning has become a significant competitive advantage for many businesses.

What are the different types of machine learning?

Classical machine learning is typically classified based on how an algorithm improves its accuracy in predictions. There are four main approaches: supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. Data scientists select the appropriate algorithm based on the type of data they want to predict. Supervised learning algorithms are effective for the following tasks:

1. Supervised learning: In this type of machine learning, data scientists provide the algorithm with labeled training data and specify the variables that they want the algorithm to analyze for correlations. Both the input and output of the algorithm are defined.
2. Unsupervised learning: This type of machine learning involves algorithms that are trained on unlabeled data. The algorithm searches through the data sets looking for meaningful connections. The data that the algorithm trains on and the predictions or recommendations it produces are predetermined.
3. Semi-supervised learning: This approach to machine learning combines elements of supervised and unsupervised learning. Data scientists may provide the algorithm with mostly labeled training data, but the model is free to explore the data on its own and develop its own understanding of the data set.
4. Reinforcement learning: Data scientists typically use reinforcement learning to teach a machine to complete a multi-step process with clearly defined rules. They program an algorithm to complete a task and provide it with positive or negative cues as it figures out how to complete the task. For the most part, the algorithm decides on its own which steps to take along the way.

How does unsupervised machine learning work?

Unsupervised machine learning algorithms do not require labeled data. They analyze unlabeled data to identify patterns that can be used to group data points into subsets. Many types of deep learning, including neural networks, are unsupervised algorithms. Unsupervised learning algorithms are effective for the following tasks:

1. Clustering: This involves dividing the dataset into groups based on similarity.
2. Anomaly detection: This involves identifying unusual data points in a dataset.
3. Association mining: This involves identifying sets of items in a dataset that frequently occur together.
4. Dimensionality reduction: This involves reducing the number of variables in a dataset.

How does semi-supervised learning work?

Semi-supervised learning involves data scientists providing an algorithm with a small amount of labeled training data. From this, the algorithm learns the dimensions of the dataset and can then apply this knowledge to new, unlabeled data. Algorithms typically perform better when they are trained on labeled datasets, but labeling data can be time-consuming and costly. Semi-supervised learning offers a compromise between the performance of supervised learning and the efficiency of unsupervised learning. Some areas where semi-supervised learning is used include:

1. Machine translation: This involves teaching algorithms to translate language based on less than a full dictionary of words.
2. Fraud detection: This involves identifying cases of fraud when only a few positive examples are available.
3. Labelling data: Algorithms trained on small datasets can learn to apply data labels to larger sets automatically.

How does reinforcement learning work?

Reinforcement learning involves programming an algorithm with a specific goal and a set of rules for achieving that goal. Data scientists also program the algorithm to seek positive rewards, which it receives when it performs an action that helps it move closer to its ultimate goal, and to avoid punishments, which it receives when it performs an action that takes it farther away from its goal. Reinforcement learning is frequently used in the following areas:

1. Robotics: This involves using reinforcement learning to teach robots how to perform tasks in the physical world.
2. Video gameplay: This involves using reinforcement learning to teach bots to play various video games.
3. Resource management: Given finite resources and a defined goal, reinforcement learning can be used to plan how to allocate resources.

Who's using machine learning and what's it used for?

Machine learning is used in a variety of applications today. A well-known example of machine learning in action is the recommendation engine that powers Facebook's news feed.

Facebook uses machine learning to personalize the delivery of each member's feed. If a member frequently stops to read posts from a particular group, the recommendation engine will show more of that group's activity earlier in the feed.

In the background, the engine is trying to reinforce established patterns in the member's online behavior. If the member changes their behavior and stops reading posts from that group in the following weeks, the news feed will adjust accordingly.

1. Customer relationship management: CRM software can use machine learning models to analyze email and prioritize responses for sales team members. More advanced systems can even suggest potentially effective responses.
2. Business intelligence: BI and analytics vendors use machine learning in their software to identify potentially important data points, patterns of data points, and anomalies.
3. Human resource information systems: HRIS systems can use machine learning models to filter through job applications and identify the most qualified candidates for an open position.
4. Self-driving cars: Machine learning algorithms can enable semi-autonomous cars to recognize partially visible objects and alert the driver.
5. Virtual assistants: Smart assistants often use a combination of supervised and unsupervised machine learning models to interpret natural speech and provide context
6. Trading: Machine learning algorithms can be used to identify patterns in financial data and make trading decisions based on these patterns. These algorithms can analyze large amounts of data quickly and make trades in real-time, allowing for the possibility of high-speed, data-driven trading strategies.
   

What are the advantages and disadvantages of machine learning?

Machine learning has been used in a wide range of applications, from predicting customer behavior to forming the operating system for self-driving cars.

One advantage of machine learning is that it can help companies understand their customers more deeply. By collecting customer data and analyzing their behaviors over time, machine learning algorithms can learn associations and help teams tailor product development and marketing efforts to customer demand.

Some companies rely on machine learning as a key part of their business model. For example, Uber uses algorithms to match drivers with riders, and Google uses machine learning to display ride advertisements in search results.

However, there are also disadvantages to using machine learning. It can be costly, as it requires skilled data scientists and expensive software infrastructure. Additionally, machine learning algorithms can be biased if they are trained on datasets that exclude certain populations or contain errors. This can lead to inaccurate models that may be discriminatory or fail to perform as expected, potentially causing regulatory and reputational harm for the company.

What is the future of machine learning?

Machine learning algorithms have been around for many years, but they have gained renewed popularity as artificial intelligence has become more prevalent. Deep learning models, in particular, are used in some of the most advanced AI applications today.

The machine learning platform market is highly competitive, with major vendors like Amazon, Google, Microsoft, and IBM all offering platform services covering a wide range of machine learning activities, such as data collection, data preparation, data classification, model building, training, and application deployment.

As machine learning becomes increasingly important for business operations and AI becomes more practical in enterprise settings, the competition among machine learning platforms is likely to intensify.

Research into deep learning and AI is focusing on developing more general-purpose applications. Currently, AI models require extensive training to produce an algorithm that is highly optimized for one specific task. Researchers are exploring ways to make these models more flexible and to develop techniques that allow machines to apply context learned from one task to future, different tasks.

What is the history of machine learning?

• 1679 - Gottfried Wilhelm Leibniz develops the system of binary code.
• 1834 - Charles Babbage conceives of a general-purpose device that could be programmed with punched cards.
  
• 1842 - Ada Lovelace writes a description of a sequence of operations for solving mathematical problems using Charles Babbage's theoretical punch-card machine, becoming the first programmer.
  
• 1847 - George Boole creates Boolean logic, a form of algebra in which all values can be reduced to the binary values of true or false.
  
• 1936 - Alan Turing proposes a universal machine that could decipher and execute a set of instructions, laying the foundations for computer science.
  
• 1952 - Arthur Samuel creates a program that helps an IBM computer get better at playing checkers the more it plays.
  
• 1959 - MADALINE, the first artificial neural network applied to a real-world problem, is used to remove echoes from phone lines.
  
• 1985 - Terry Sejnowski and Charles Rosenberg's artificial neural network teaches itself to correctly pronounce 20,000 words in one week.
  
• 1997 - IBM's Deep Blue defeats chess grandmaster Garry Kasparov.
  
• 1999 - A CAD prototype intelligent workstation reviews 22,000 mammograms and detects cancer 52% more accurately than radiologists.
  
• 2006 - Geoffrey Hinton coins the term "deep learning" to describe neural net research.
  
• 2012 - An unsupervised neural network created by Google learns to recognize cats in YouTube videos with 74.8% accuracy.
  
• 2014 - A chatbot passes the Turing Test by convincing 33% of human judges that it is a Ukrainian teen named Eugene Goostman.
  
• 2014 - Google's AlphaGo defeats the human champion in Go, the most difficult board game in the world.
  
• 2016 - LipNet, DeepMind's artificial intelligence system, accurately lip-reads words in video with 93.4% accuracy.
  
• 2019 - Amazon controls 70% of the market share for virtual assistants in the U.S.