What is Machine Learning?
Consider this journey essentially similar to a thoroughly crafted meal: we’ll begin with appetizers that serve as a primer on what Machine Learning entails, including its sub-fields and various types, painting a clear picture of its extensive journey which is much similar pattern to how we dine.
Introduction
From there, we’ll transition to the main course, where we’ll delve into the real-world applications of machine learning across diverse sectors such as HealthCare, Astrophysics concepts also called PINNs, Law, Finance, and more where we haven’t thought. And just like any well-structured meal, we’ll top it off with a delectable dessert — code implementation and proof results — to satisfy your intellectual palate. Let’s embark on this journey together!
Today’s Special Menu:
What is Machine Learning? How is “machine” learning? Machine Learning as a system? Types of Machine Learning Systems?
What is Machine Learning?
“A program or system that trains a model from input data. The trained model can make useful predictions from new (never-before-seen) data drawn from the same distribution as the one used to train the model” by Google Fundamentals.
“The field of study that gives computers the ability to learn without explicitly being programmed” by Arthur Samuel.
“Machine Learning is a topic that solves and automates repetitive tasks with integration and adapts human behavior of observing patterns.” by Me.
Before this massive crowd out of Machine Learning theories, concepts, applications, and modifications, we perceived its presence although we didn’t know WHY? because it wasn't statistically enhanced like today. To explain this let me take you into the kitchen concept. Core ingredients used in the kitchen, for instance, mixed spices added to vegetables along with some heat and water make it as a whole dish. This is the very basic process that every kitchen follows from a 1-star rating to 7 stars, maybe the way(method) or extra utensils or techniques might be different but the fundamental approach remains the same. Let me start with spices used in an Indian kitchen or “Masalas” like turmeric, red chili powder, coriander powder, or cumin powder which is the same as the DATA we use in ML algorithms. Data can be in any form and will be discussed later. Now you might be getting some feel for the need for Data.
But some unanswered questions may arise like WHY(Why we need data for ML Algo) and WHERE(Where to fit the data or where we can get the data or mainly where is the data ). Let me cook…….
Mixed Species: Data, Vegetables: Model selection and Training, Heat: Feature Engineering, Water: Optimization and Recursion. Now maybe the above paragraph would make sense and if not you have still many chances.
Real-Life Example: This might be detailed [intermediate can skip and suggest to directly jump to the next dish on today’s menu or read this https://www.linkedin.com/pulse/dream-90s-fico-delivers-xai-fraud-detection-models-scott-zoldi/ ].
Machine learning has been extensively used in credit scoring to assess the creditworthiness of individuals and businesses. The first credit score was introduced in 1989, but attempts to standardize this process began as far back as 1841. Between the 2000s and 2010s, the use of machine learning in credit scoring became more practical. Credit scoring involves predicting the likelihood of a borrower defaulting on a loan or failing to make timely payments based on various factors such as income, credit history, debt levels, and other relevant variables. HOW DID THEY COOKED THIS?
Steps:
1: Data Collection and Processing:
The first step in credit scoring is to gather relevant data about the borrower. This typically includes information such as income, employment history, credit history (credit score, number of open accounts, payment history, etc.), debt-to-income ratio, loan amount, loan purpose, and demographic information.
Once the data is collected, it undergoes preprocessing, which involves cleaning, transforming, and standardizing the data. This step may include handling missing values, encoding categorical variables, scaling numerical features, and removing outliers. (I will cover this later).
2: Feature Engineering: For now, think of it as like in cooking, where the selection of ingredients is crucial for the taste and quality of the dish, feature engineering involves choosing the right data variables (ingredients) that are relevant and meaningful for the machine learning model. Selecting the right features can greatly impact the performance and accuracy of the model. (I will cover this later)
3: Model Selection: Let’s make it sound simple, just as a cook selects a recipe based on factors such as available ingredients, dietary restrictions, and desired outcome, model selection involves choosing the appropriate machine learning algorithm based on factors such as the nature of the data, the problem at hand, computational resources, and the desired outcome (e.g., classification, regression, clustering). (I will cover this later)
4: Model Training, Evaluation, and Validation: These 3 are distinct yet interconnected stages in the machine learning process. But just to make you feel how simple the basic outline structure of Machine Learning is my primary goal. (I will cover this later).
5: Interpretability and Fairness: Just as a chef may disclose the ingredients and cooking methods used in a dish to ensure transparency, interpretability in machine learning involves making the model’s decisions understandable and transparent. Users should be able to understand how the model arrives at its predictions, much like understanding the steps taken to prepare a meal. (I will cover this later too)
How is “Machine” learning? / What makes Learning possible for Machines?
There are 4 ingredients to make this concept possible- Data, Algorithm, Generalization, and Statistics. We know that Researchers won’t sit idle besides proving their work every single day. Therefore, the idea of machines learning from the data, dates back to the mid-20th century. In the 1950s and 1960s, researchers explored the possibility of creating artificial intelligence (AI) by programming computers to mimic human problem solving BUT, lacked computational power and the complexity of real-world problems. As We know, Researchers won’t sit idle and therefore researchers began to develop algorithms that could analyze data and learn patterns from examples. This started with improving a simple learning board evaluation algorithm. This marked the beginning of modern machine learning. The explosion of digital data in the late 20th and early 21st centuries provided the fuel for machine learning algorithms. HOW THEY APPROACHED WHAT ALGORITHM TO MAKE? Researchers developed and refined a wide range of machine learning algorithms and techniques, each suited to different types of data and tasks. These include supervised learning algorithms (e.g., regression, classification), unsupervised learning algorithms (e.g., clustering, dimensionality reduction), and reinforcement learning algorithms, among others.
Machine Learning as a System?
The systems can also be said to be init types which are characterized based on the TYPE of problem, AVAILABILITY of resources for the problem, DATA TYPE given for the problem, END RESULT, and more. In detail…
Type- Numerical, Categorical, Text, Image, Audio, Video, Time Series, Biological data, Panel, Streaming, and more, which are still under exploration. As chef tries different recipes similarly Researchers won’t sit idle.
Availability- Computational Resources, Power, Memory, Storage, and size of data.
Data Type- The type of data significantly influences the choice of preprocessing techniques, feature engineering methods, and modeling algorithms in choosing machine learning flow and goal.
End Result- Prediction, Classification, Regression, Anomaly Detection, Recommendation, Sequence Ranking and many more.
Answers to many more will be discussed in Chapter 3/4……..
Types of Machine Learning Systems?
Let’s take this dish bite by bite, delving into the most important types that encapsulate almost all discovered Machine Learning concepts.
1: Supervised Learning
2: Unsupervised Learning
3: Reinforcement Learning
These three types represent the foundational pillars of machine learning, each with its own unique approach and applications.
The next topping serve includes Types in Machine Learning-> Supervised Learning in depth, which is “Hello World” in Machine Learning.
References:
