Machine Learning in a nutshell
Before we talk about machine learning, we better talk about Artificial Intelligence first. If we say the term AI, some people will see AI as a robot, or a machine that works and behaves similarly to humans. If you have watched these movies: Westworld, The Matrix, Wall-E, Chappie, you will get some ideas on AI.
Artificial Intelligence is exactly the simulation of human intelligence process by machines, especially by computer process or system. Expert systems, natural language processing, speech recognition, computer vision and machine learning are applications of AI.
There are three different types of AI: Artificial Narrow Intelligence (ANI), Artificial General Intelligence (AGI), and Artificial Super Intelligence (ASI).
Weak AI (ANI) is programmed to perform a single task. For example, checking the weather, being able to play chess, or analyzing raw data to write journalistic reports.
Strong AI (AGI) can successfully perform any intellectual task that a human being can. It has the abilities to reason, solve problems, make judgments under uncertainty, plan, learn, integrate prior knowledge in decision-making, and be innovative, imaginative, and creative.
Super AI (ASI) will surpass human intelligence in all aspects from creativity, to general wisdom, to problem-solving.
AI is useful in a variety of industries like healthcare, retails, banking and finance, agriculture, and manufacturing.
Now then, let me explain briefly its applications.
Expert system
An expert system is a computer system that emulates the decision-making ability of a human expert.
Application Scenarios
Design Domain, Medical Domain, Monitoring System, Process Control System, Knowledge Domain, Finance/Commerce
Natural Language Processing (NLP)
NLP helps computers understand, interpret and manipulate human language. NLP draws from many disciplines, including computer science and computational linguistics, in its pursuit to fill the gap between human communication and computer understanding.
Applications
- Language translation applications such as Google Translate
- Word Processors such as Microsoft Word and Grammarly employ NLP to check the grammatical accuracy of texts.
- Interactive Voice Response (IVR) applications are used in call centers to respond to certain users’ requests.
- Personal assistant applications such as OK Google, Siri, Cortana, and Alexa.
Automated Speech Recognition
ASR allows human beings to use their voices to speak with a computer interface in a way that, in its most sophisticated variations, resembles normal human conversation.
Computer Vision
CV interprets and understands the visual world. Using digital images from cameras and videos and deep learning models, machines can accurately identify and classify objects. Object detection, face recognition, Optical Character Recognition (OCR) are parts of computer vision.
That’s enough about AI.
Let’s jump to Machine Learning.
What exactly Machine learning is
Machine learning is an application of artificial intelligence (AI) that provides systems the ability to automatically learn and improve from experience without being explicitly programmed. Machine learning focuses on the development of computer programs that can access data and use it learn for themselves.
If we talk about Machine learning or AI, the most important factor is Data. Data can be collected anywhere every day. The reason behind AI/ML has become more popular today is increased data volumes, advanced algorithms, and improvements in computing power and storage.
Imagine that you show a cat picture to a baby, how a baby can know what exactly that animal is. As human nature, we observe, learn and use the skill that we know to classify the objects like cats.
The machine also does the same as human, the data, cat, is prepared and fed to a machine learning model as input and it learns the features like body size, ears, eyes, paws, from the input data, then the model predicts the object or here, a cat label. To know the output, we have to teach the model, what we called training the machine learning model.
There are three main methods of machine learning: Supervised Learning, Unsupervised Learning, and Reinforcement Learning. Here, we call the data what we will be training as features, and the output/result that we would like to get is called target.
In supervised learning, features are given with labels and the model knows what classes to be output on known data. On the other hand, it can make predictions for unknown data that we obtain later in the future. Classification and Regression fall into supervised learning.
In unsupervised learning, features are given without labels, and the model does not have known results. It identifies patterns in the data and helps humans make decisions. Clustering and Dimensionality Reduction fall into unsupervised learning.
In reinforcement learning, the algorithm or agent learns from the environment by interacting with it. RL is closest to how a human learns. Based on the actions, rewards, positive or negative is given to the agent.
Cool, I believe you are clear about what machine learning is. We are not diving deeper into this article. Despite that, I want to show the general process flow of machine learning which is useful in most scenarios.
The 7 Steps of Machine Learning
- Data Collection — This step is the core if you do machine learning. Without data, we can do nothing. Like I explained earlier, there are a variety of sources that we can collect data, for example, websites, social media, your organization, or your company. You will later understand that the data sometimes can be collected ourselves, sometimes use public or request published dataset.
- Data Preparation — We have to wrangle and prepare data to the format that we want. We do some processes in a single step like data cleansing, data transformation, also splitting the dataset into training, validation, and testing to get ready for machine learning model training.
- Choose a Model — Choose the model that is suitable for our business scenarios. There is no right or wrong answer for choosing the best model. But, we can figure out the pros and cons of each model, which will support choosing the model.
- Train the Model — Once the model has been chosen or decided, we feed the prepared training dataset to the model, and the model starts learning the features of them, what we called model training.
- Evaluate the Model — After the model is trained on the training dataset, it is required to validate or evaluate the performance of the model on the validation dataset whether it predicts the targets correctly or not.
- Parameter Tuning — We cannot assure that the model will be good enough for the first time training. It requires hyper-parameter tuning, which searches or choosing the optimal parameters, till it meets our predefined accuracy or precision score.
- Make Predictions — To assure the model's accuracy, the model has to be tested on unknown or unseen data, what we call a testing dataset, which has never been seen by the model before. You can assume that data is excluded on both the training and validation datasets.
Hope this makes you understand machine learning and its process steps which is quite useful when you want to start learning machine learning.
Feel free to check out our LinkedIn Page, https://www.linkedin.com/company/ai-leap
