04 Dec 23

Demystifying Machine Learning

Max Collins

Machine Learning (ML) often seems like a magical black box, you just give it an input and it knows the output instantly, but how does it actually work? In this blog, I will attempt to demystify Machine Learning, breaking it down into understandable concepts.


What is Machine Learning?






At its core, Machine Learning is a way for computers to learn and make decisions based on data. Unlike traditional programming where exact instructions are given, Machine Learning systems learn patterns from data and make predictions or decisions without being explicitly programmed for the task.

The process begins with a dataset of examples to train the Machine Learning model. These examples consist of input data. (Features) and corresponding output labels (targets). The model learns to predict the outputs from the inputs by identifying patterns and relationships within the data.



Types of Machine Learning

There are three main types of Machine Learning: supervised learning, unsupervised learning, and reinforcement learning.





The Building Blocks of Machine Learning Models

Behind the scenes, Machine Learning models are powered by algorithms—mathematical procedures that enable the system to learn from data. Some common Machine learning algorithms include:






Each algorithm has its strengths and weaknesses, and the choice of algorithm depends on the nature of the data and the problem being addressed.



How are they trained?

To learn from the training dataset, ML models iteratively refine themselves using optimisation techniques such as gradient descent, stochastic gradient descent, and variants like Adam and RMSprop. Don’t worry too much about the specifics of these, These techniques essentially aim to find the best parameters for the model to minimise the difference between their predictions and the actual outcomes in the training set.


Once the model has learnt from the training data, it is put to the test using a separate dataset that it has never seen before. It is important that this is new data so we can be sure the model hasn’t overfit the dataset. Overfitting in Machine Learning can be thought of as a student memorising specific answers to a test rather than understanding the underlying principles. It may perform well on that specific test, but as soon as you give it a new one it will fail. Common metrics, such as precision, recall, and accuracy, are used to measure the model’s performance on the testing dataset.



Challenges and Considerations

While ML can be very useful for a variety of tasks, it’s not without its challenges. Here are some common issues:






Real-World Applications

Here are some practical applications of Machine Learning that you encounter in your daily life:







In conclusion, Machine Learning is not an elusive magic trick, but a set of mathematical algorithms trained on large datasets to understand patterns. By understanding the basics of how Machine Learning works and its real-world applications, you can appreciate its impact on various aspects of your life.

I hope this overview has provided useful insights into ML and a better understanding of what’s happening behind the scenes next time you use something like ChatGPT.


Interested in joining our diverse team? Find out more about the Rockborne graduate programme here.

Apply Today