Table of contents
- What is Artificial Intelligence
- What is Machine Learning
- What is Deep Learning
- What is Supervised Learning
- What is Unsupervised Learning
- What is Semi-supervised Learning
- What is Reinforcement Learning
- What is Classification
- What is Regression
- What is Clustering
- What is Backpropagation
- What is Natural Language Processing
- What is Neural Network
- What is Computer Vision
What is Artificial Intelligence
A sub-field of computer science. It is an intelligence demonstrated by machines in contrast to the natural intelligence displayed by humans.
It is often used to describe machines (or computers) that mimic "cognitive" functions that humans associate with human mind, such as "learning".
It includes,
- understanding human speech
- competing at the highest level in strategic game systems, such as chess.
- Autonomously operating cars.
- Intelligent routing in CDNs
- Military simulations and etc.
The most popular sub-field of AI achieving this is Machine Learning.
What is Machine Learning
It's a subset of Artificial Intelligence. It's a set of algorithms that allow computers to "learn" from data without being explicitly programmed. It uses statistical methods to enable machines to improve using a large amount of data, like annotated pictures of different cats. The machine can then determine whether there's a cat in a new unseen image by comparing what it has seen before, just like humans.
This is just an example of classification, where we only say if there is a cat or not in the picture. There’s a lot of different ways computers “learn” from data to achieve an end goal decided by the developer. Machines “learn” from data, either by using a supervised, semi-supervised or unsupervised method. Which are all under a subset of machine learning called Deep Learning.
What is Deep Learning
Also known as deep structured learning or hierarchical learning.
Is a subset of artificial intelligence and machine learning. The specification of deep learning is that the neural network has way more layers than any other machine learning subset, it is way “deeper”. This is the reason why it required a large amount of training data. Resulting in extremely powerful learning models capable of processing data in new ways. But this takes a lot of time and power to train for a machine.
It is basically connecting a lot of “neurons” together, just like the brain, assigning weights to each feature. It can then decide the correct output from any input it receives, connecting the dots together. Like a detective who links all the clues to find the guilty person.
Learning these weights can be done using supervised, semi-supervised, unsupervised and reinforcement learning.
What is Supervised Learning
Is one of the many ways machine learning algorithms learn.
It learns a function that maps an input to an output based on example input-output pairs.
It infers a function from labeled training data consisting of a set of training examples. In supervised learning, each example is a pair consisting of an input object (typically a vector) and a desired output value (also called the supervisory signal).
This is the most accurate way to train a machine learning model at the moment. But it requires a lot of annotated data for its training. Which is frequently really difficult to get. It is mostly used for regression and classification problems.
The problem of this method consists in generalising to new, unseen data. So the model does not only succeed on the data seen in its training. Both the lack of data and the generalisation problem pushed the development of other types of learning methods.
Like Unsupervised and semi-supervised learning.
What is Unsupervised Learning
Is a type of self-organized learning that helps find previously unknown patterns in a data set without pre-existing labels. It is also known as self-organization and allows modeling probability densities of given inputs.
It infers a function from unlabeled training data consisting of a set of training examples. Two of the main methods used in unsupervised learning are principal component and cluster analysis.
Cluster analysis is used in unsupervised learning to group, or segment, datasets with shared attributes in order to extrapolate algorithmic relationships. Instead of responding to feedback, cluster analysis identifies commonalities in the data and reacts based on the presence or absence of such commonalities in each new piece of data. This approach helps detect anomalous data points that do not fit into either group.
It is one of the main three categories of machine learning methods used, along with supervised, reinforcement learning and Semi-supervised learning.
What is Semi-supervised Learning
Is a hybrid between supervised and unsupervised learning. It uses both labeled and unlabeled data during training.
It is used because getting labeled data is hard in many ways, but it can produce considerable improvement in learning accuracy.
Since acquisition of unlabeled data is relatively inexpensive, adding a small portion of expensive labeled data to it is overall cheap and produces extremely great results in comparison with only unlabeled data.
Intuitively, the learning problem can be seen as an exam and labeled data as sample problems that the teacher solves for the class as an aid in solving another set of problems where you don’t have access to the answer, in this case the unlabeled data.
It is good practice for the exam since you have seen examples on how to do them and can compare them.
What is Reinforcement Learning
Is an area of Machine learning concerned with how software agents ought to take actions in an environment in order to maximize some notion of cumulative reward.
It differs from supervised learning in not needing labeled input/output pairs to be presented, and in not needing suboptimal actions to be explicitly corrected.
Instead the focus is about taking suitable action to maximize reward in a particular situation. So reinforcement learning is all about making decisions sequentially.
Let’s explain it with an example, a chess game. As said previously, we have an agent and an environment. The agent would be the player and the environment is the chess game. Then, reinforcement learning would be a conversation between the two, being on one side the actions from the agent to the environment, a move from the player. And on the other side, states from the environment to the agent; results of the move. The environment will also tell the agent if the move was bad or not.
In our example, the agent will know its actions were good, if he wins. It will then be positively rewarded and the agent will tend to use the same type of actions. In this case the reinforcement is positive, but it can also be negative as in punishing the agent for doing bad actions.
After multiple iterations, the agent will be more and more confident and have great results. A great example of this is AlphaZero by Google.
What is Classification
Is an instance of supervised learning.
A classification model attempts to identify to which of a set of categories a new observation belongs, based on a training set of data containing observations whose category membership is known.
For example, we use separate dustbins for different types of waste: one for paper waste, one for plastic waste, one for dry waste, one for wet waste and so on. These waste types will be our categories.
If you have an old cup of coffee made of plastic, you'll put it in the plastic dustbin because you know it has the same property; it's made of plastic. What you're basically doing here is classifying the waste into different categories and choosing one; plastic. In short, classification is the process of assigning a 'class label' to a particular item.
There are multiple classification algorithms like linear classifiers, support vector machines, kernel estimation, decision trees and much more.
What is Regression
Is an instance of supervised learning.
It is a reliable method of identifying which variables have an impact on a topic of interest.
It is used for estimating the relationships between a dependent variable (often called the 'outcome variable') and one or more independent variables (often called 'predictors', 'covariates', or 'features').
The process of performing a regression allows you to confidently determine which factors matter most, which factors can be ignored, and how these factors influence each other.
Regression is used to identify a tendency between variables and deduce an output from it. There are multiple regression algorithms like linear regression, polynomial regression, ElasticNet regression and much more.
What is Clustering
Is an instance of unsupervised learning.
It is the task of grouping a set of objects in such a way that objects in the same group (called a cluster) are more similar to each other than to those in other groups.
In machine learning, we often group examples as a first step to understand a subject (data set) in a machine learning system. Grouping unlabeled examples is called clustering.
Before you can group similar examples together, you first need to find similar examples. You can measure similarity between examples by combining the examples feature data into a metric, called a similarity measure.
Let's see a simple comparison example, Imagine you're in a supermarket in another country and see a food you've never seen before. The label is not in your language, but it's just between what seem to be apples and pears. You'll then assume it's a fruit, and probably similar to apples and pears even though you can't read the "fruit" label section in that unknown language.
What you did is you made a cluster named "fruits" and you classified the unknown food to it. You used proximity and shape as input variables to deduce a common output with the apples and pears separating them from the rest of the food in the supermarket.
In short, grouping unlabeled examples is called clustering. As the examples are unlabeled, clustering relies on unsupervised machine learning. If the examples are labeled, then clustering becomes classification. There are multiple clustering algorithms like K-means, mean-shift, gaussian mixture and much more.
What is Backpropagation
Is a supervised learning algorithm used for training a deep learning model. This learning method is the most popular at the moment because it makes possible the use of powerful calculation methods with low computation time.
In a deep learning model, we use the parameters to calculate the usage proportion of each neurons to get an expected output. Training a model means teaching it which parameters to use to successfully get the right output. With the use of backpropagation, the model will change its parameters (called weights) to achieve the expected output. It changes its parameters by calculating the difference between the expected output and the actual output, called the error.
The backpropagation algorithm is simple to understand. At first, we send an example, like a picture of a cat to the model. Since the model is not trained yet, we don't know which weights to use and not use. We have to initialise the weights to some random value.
But now we get an output 'dog' instead of the expected output 'cat'. The model calculates the difference between the results for 'cat' and the results that got us 'dog'. This is the error. It then propagates this error backward to the model updating all the weights in proportion to the error.
After doing so, another example is fed to the model and we repeat the process all over again, until the error is minimized and the model is trained. Once the error become minimum, your model is ready to work with unseen data.
The calculation behind this algorithm are quite complex and requires techniques like gradient descent, mean squared error and much more.
What is Natural Language Processing
Also known as NLP
Is a sub-field of linguistics, computer science and artificial intelligence concerned with the interactions between computers and human (natural) languages.
The ultimate objective of NLP is to read, decipher, understand, and make sense of the human languages in a manner that is valuable. Most NLP techniques rely on machine learning to derive meaning from human languages. NLP is the driving force behind the personal assistants, word processors like microsoft word, language translation applications and much more. It is one of the biggest subfields of AI mainly because you use it everyday; with Siri, google translate, microsoft word, etc.
Humans can easily master a language, the ambiguity and imprecise characteristics of the natural languages are what make NLP difficult for machines to implement.
These challenges in natural language processing frequently involve speech recognition, natural language understanding, and natural language generation.
What is Neural Network
A neural network is a network of neurons, or in a modern sense, an artificial neural network, composed of artificial neurons. It mimics the brain's neurons using nodes, connections and weights to find an output relative to a specific input. A neural network is based on a collection of connected units or nodes called artificial neurons, which loosely model the neurons in a biological brain. Each connection, like the synapses in a biological brain, can transmit a signal to other neurons. An artificial neuron that receives a signal, then processes it, can send an other signal to the neurons connected to it. Such systems "learn" to perform tasks by considering examples, generally without being programmed with task-specific rules.
What is Computer Vision
Is a subfield of artificial intelligence that trains computers to interpret and understand the visual world. It includes methods for acquiring, processing, analysing and understanding digital images. The problem of computer vision appears simple because it is trivially solved by people, even very young children.
It is whole other deal for a computer.
It is largely remains an unsolved problem based both on the limited understanding of biological vision and because of the complexity of vision perception in a dynamic and nearly infinitely varying physical world.
The study of computer vision and neuroscience is closely linked together to understand both how we see and how do we know what we see. In short, the goal of computer vision is to understand the content of digital images. Typically, this involves developing methods that attempt to reproduce the capability of human vision.
Understanding the content of digital images may involve extracting a description from the image, which may be an object, a text description, a three-dimensional model, and so on. It is a huge subfield of AI used across industries to enhance the consumer experience, reduce costs, increase security and much more.
You've come to the end. Thanks for reading !
Note : The above list of terminologies are not finished yet, I'll keep updating the lists as I grow myself into this field. You can bookmark this blog if you like to learn more terms or just for quick go through whenever you need in future.
Until then
Keep Building Yourself
Cover Photo by Markus Winkler on Unsplash