AI Models and Algorithms

- Algorithms

An algorithm is simply a set of steps used to accomplish a specific task. They are the building blocks of programming that allow devices such as computers, smartphones, and websites to operate and make decisions.

Algorithms have been around for a long time before the general public notices them. The term is simple: an algorithm is just any step-by-step procedure for accomplishing some task, from making your morning coffee to performing heart surgery. Algorithms are used in almost everything a computer does. 

But when algorithms start taking over tasks that used to require human judgment, allowing machines to think and make decisions like humans, they become harder to ignore. For example, deciding which criminal defendants get bail, screening job applications, and prioritizing stories in news feeds. 

Since the development of complex artificial intelligence (AI) algorithms, it has been possible to achieve this by creating machines and robots that are used in a wide range of fields, including agriculture, healthcare, robotics, marketing, business analytics, and more. Over time, the potential for AI to mimic and surpass the capabilities of the human mind grows exponentially.

- AI Models

An artificial intelligence (AI) model is a program or algorithm that relies on training data to recognize patterns and make predictions or decisions. The more data points an AI model receives, the more accurate its data analysis and predictions will be. 

AI models rely on computer vision, natural language processing, and machine learning to identify different patterns. AI models also use decision-making algorithms to learn from training, collect and review data points, and ultimately apply learning to achieve predefined goals. 

AI models are very good at solving complex problems with large amounts of data. As a result, they are able to accurately solve complex problems with very high accuracy.

- Machine Learning Algorithms

Typically, an algorithm takes some input and uses mathematics and logic to produce an output. In stark contrast, AI algorithms take a combination of both inputs and outputs in order to "learn" data and produce output when given new inputs. 

This process of letting machines learn from data is what we call machine learning (ML). ML is a subfield of AI where we try to bring AI into the equation by learning from input data. 

Artificial intelligence now means the so-called "second wave artificial intelligence" or "narrow artificial intelligence". This is a very different project, focused on machine learning (ML). 

The idea is to build systems that can mimic human behavior without necessarily understanding it. The way you train an algorithm is similar to how a psychologist trains a pigeon to distinguish a picture of Charlie Brown from a picture of Lucy. 

You give it a bunch of data — posts that Facebook users have engaged with, comments that human commenters have classified as toxic or benign, messages marked as spam or not spam, and so on. 

The algorithm considers thousands or millions of factors until it figures out on its own how to distinguish categories or predict which posts or videos someone will click on. At that point you can put it in the world. 

Machines can learn in different ways depending on the dataset and the problem being solved. Machine learning can be done in the following ways: supervised learning, unsupervised learning, reinforcement learning, and ensemble learning.   

[Murren, Switzerland - Civil Engineeriing Discoveries]

- Machine Learning Models Vs. Artificial Intelligence Models

Many people mistakenly confuse machine learning and artificial intelligence. This may be because machine learning is a subset of artificial intelligence. However, there are key differences between the two that you should be aware of. 

As we defined it earlier, artificial intelligence involves the creation of machines that simulate human thought, intelligence, and behavior. 

Machine learning, on the other hand, strives to provide machines with the ability to learn on their own from experience and lessons without the need for explicit programming. 

All machine learning models are artificial intelligence models, but not all artificial intelligence models are necessarily machine learning models. This is an important distinction that will help you understand this topic in more detail. 

Machine learning models are an important part of human intelligence, which is to learn things and predict future results based on past experiences and lessons. Likewise, AI models learn based on annotated data during the learning phase.

- Search Algorithms and Search Space in AI

Search algorithms are algorithms used in AI to find the best solution to a problem. They work by defining the problem and conducting search operations. 

Some properties of search algorithms include: Optimality, Completeness, Time complexity, Space complexity. 

Whatever you do, start with a search! AI can solve these everyday problems. To some extent, many AI problems can be modeled as search problems, where the task is to reach a goal from an initial state through state transition rules. 

Therefore, the search space is defined as a graph (or tree), and the goal is to reach the goal from the initial state through the shortest path, including cost, length, combination of the two, etc.

A given problem can have one or more solutions, depending on the scenario, since there can be multiple ways to solve the problem. Think about how you approach a problem. Now, it's like playing a game (imagine your favorite game, chess, poker, whatever...). 

In most games of this type, at a given point in time, you can take multiple actions and choose the one that gives you the best outcome. In this case, there is no one right solution, but one possible best solution, depending on what you want to achieve. Also, depending on the game strategy you choose, there are various ways to approach this problem. 

- The Rational Agents For AI

Rational agents of AI approach the problem in a similar way. It must search the solution space to provide the best results. This makes search algorithms important in AI research. 

As to what is considered the best outcome and why one solution is better than another, is something we program into AI. We will see how AI searches for a solution to a given problem.

- Types Of Problems Solved Using AI Algorithms

AI algorithms can be used to solve different types of problems. Here is an overview and brief overview of machine learning problem categories: 

• Classification: Classification is the act of dividing the dependent variable (the variable we are trying to predict) into categories and then predicting the category for a given input. It belongs to the category of supervised machine learning, and the data set first needs to have classes. So classification comes into play wherever we need to predict an outcome from a fixed, predefined set of outcomes. Classification uses a range of algorithms, some of which are listed below: Naive Bayes, Decision Trees, Random Forests, Logistic Regression, Support Vector Machines, K-Nearest, Neighbors.
• Regression: In case of regression problems, the output is a continuous quantity. This means that we can use regression algorithms in cases where the target variable is a continuous variable. It belongs to the category of supervised machine learning, and the data set first needs to be labeled.
• Clustering: The basic idea behind clustering is to assign inputs into two or more clusters based on feature similarity. It falls under the category of unsupervised machine learning, where an algorithm learns patterns and useful insights from data without any guidance (labeled dataset). For example, unsupervised learning algorithms such as K-Means clustering can be used to cluster audiences into similar groups based on interests, age, geographic location, etc.

Algorithms in each category perform essentially the same task of predicting an output given an unknown input, however, data is the key driver when choosing the right algorithm.

[More to come ...]