Machine Learning – A Gentle Approach part 1

AI vs ML:

Machine Learning focuses on teaching machines how to perform a task, by giving them a lot of data samples enabling them to identify patterns in data. After capturing these patterns, machines can finally guess an outcome (ex. predict a value or a class) of a newly introduced data.

The importance of Machine learning lies in letting the machine learn the patterns and then make decisions without being explicitly programmed what to do for every scenario.

AI, on the other hand is a broader field that encompasses the development of computer systems that use algorithms (ML or other techniques) to learn and adapt from data.

How does machine learning work?

Performing machine learning involves a series of steps:

1.??? Data collection: Machine learning starts with gathering data from various sources, such as photos and tabular data (ex. excel). The raw data containing input columns, also known as features, is then organized, and prepared for use as training data for the computer to learn from.

2.??? Data preparation:?This step includes cleaning the data and feature engineering (ex. selecting best features then scaling and normalizing data points)

3.??? Choosing and training the model:?Depending on the task at hand, engineers choose a suitable machine learning model and start the training process. The computer model automatically learns from the data by searching for patterns and adjusting its internal settings to optimize a specific objective. It essentially teaches itself to recognize relationships and make predictions based on the patterns it discovers.

4.??? Model optimization:?Engineers can enhance the model’s performance by adjusting its parameters or settings. By experimenting with various configurations, Engineers try to optimize the model according to the specific objective at hand (make precise predictions or identify meaningful patterns in the data).

5.??? Model evaluation:?Once the training is over, engineers need to check how well the model performs. One way to do this, is by evaluating its ability to generalize data (how well the model performs on new/unseen data).

6.??? Model deployment:?After the model has been trained and evaluated, the model is deployed to:

• ?????? Fulfill tasks in real-time (ex. User Interface for Stock Market Predictions)
• ?????? Dynamic learning by continuously adding new data samples to training dataset pool, aiming to represent the total data more accurately and improve performance.

There are Four main ML types:

Data types:

1.???? Categorical: Non-Numerical values in the form of categories or labels.

• ?????? Ordinal: In Order, but the differences between categories may not be uniform (ex. low, high)
• ?????? Nominal: No inherent order (ex. red, green, blue)

2.???? Quantitative: Numerical values that can be ordered in a meaningful way.

• ?????? Discrete: Integers within a small range (usually whole digits ex. 1, 2, 3)
• ?????? Continuous: All real numbers (ex. 3.14, 9.8)

?

Machine Learning models:

1.???? Supervised Learning models:

• ???? Classification: Teaching a computer to recognize and label things into pre-defined groups based on common patterns in the trained features.
• ???? Regression: Making predictions or estimating values. It's like trying to find a mathematical rule that helps you predict an outcome.

?

2.???? Unsupervised Learning models:

• ???? Clustering: Trying to group data into clusters based on common patterns in the trained data samples. Then each cluster is examined to determine what it represents.
• ???? Dimensionality Reduction: Reduce the number of non-significant features to optimize performance and it also makes it easier to visualize the data's structure and relationships among data points.

?

3.???? Semi-Supervised Learning models:

• ???? Classification & Regression: Semi-Supervised Learning is used when a combination of labelled and unlabeled data is available. The task is usually either classification or regression.

?

4.???? Reinforcement Learning models:

• ???? Control: A learning agent interacts with an environment through trial and error. The agent receives feedback in the form of rewards or penalties based on its actions in different states. The agent's goal is to learn a policy (a strategy) that maximizes its cumulative reward over time.

Data needed for each model:

1.??? Supervised Learning data consists of both: features (input columns) and labels (output column/s)

2.???? Unsupervised Learning data consists only of features (input columns)

3.???? Semi-Supervised Learning data: usually consists of small labelled (Supervised) dataset and large unlabeled (Unsupervised) dataset.

4.???? Reinforcement Learning data: typically presented as a sequence of states, actions, rewards, and possible next states.

?

Kind Note:

If you're intrigued by the applications of ML and AI in the corporate world, stay with us. Gaining a solid grasp of the fundamentals is an investment of your time that promises substantial returns.

In the upcoming parts of this series, we'll delve deeper into the various aspects we've touched on.

Get ready to unlock hidden strategies and discover the insider's guide to achieving peak performance.

Q: What's one concept from this article that you found particularly intriguing or confusing?

Thank you for reading! Hope to see you again in our newsletter.

Special thanks to Analytics Vidhya for the inspiration behind this part of the series.

#TechMinds #AI, #MachineLearning, #DataScience, #TechEducation, #LearnML, #CareerAdvancement, #AIforBeginners, #MLFundamentals, #ArtificialIntelligence, #TechInsights, #MLSeries, #TechAdvantage