What is an Overfitting?
When a model memorises training examples too precisely and fails to generalise to new data.
Overfitting explained in plain English
Overfitting happens when a system memorises its training examples too precisely and fails on new data. It performs brilliantly on what it has seen before but struggles when faced with slightly different situations.
Good models balance fitting the training data with remaining flexible enough for the real world.
Analogy
Overfitting is like a student who memorises every past exam paper word for word but cannot answer a question phrased slightly differently. They mastered the examples, not the subject.
Example
An hiring model trained on past decisions may overfit to historical bias and reject qualified candidates who look different from previous hires.
How is Overfitting used?
A language model that quotes training data verbatim instead of generalising. A photo classifier that only recognises images taken in one specific lighting condition. Any AI that aces tests but fails in the real world.
Common misconceptions about Overfitting
People also read
- Artificial Intelligence
Any system that performs tasks requiring human judgment — understanding language, recognising images, making decisions, or solving problems — by finding patterns in data rather than thinking or feeling.
- Backpropagation
The process that tells a neural network which internal settings caused an error and how to adjust them, working backwards through layers.
- Classification
The task of sorting inputs into predefined categories — choosing a label rather than producing a number.
- Embedding
A numerical representation of text, images, or other data that captures semantic meaning.
- Fine-tuning
The process of further training a pre-trained AI model on specialised data to improve performance on specific tasks.
- Gradient Descent
The method by which a model gradually improves by making small adjustments after each mistake, moving toward better performance.
- Inference
The phase when a trained model is actually used — taking new input and producing a prediction or response.
- Machine Learning
A way for computers to learn from examples instead of being given exact rules — by finding patterns in labelled or unlabelled data.
- Neural Network
A layered system that processes information in stages, with each layer detecting slightly more complex patterns than the last.
- Supervised Learning
Training a system using examples where the correct answer is already known, so it learns to map inputs to labels.