My Knowledge Wiki
  • Introduction
  • My Stack
  • Sharing
  • Work
    • How I Want to Work
    • Hiring Process
  • Apps & Tools
    • Raycast
      • Extensions
    • Visual Studio Code
      • Theme
      • Extensions
    • Google Chrome
      • Theme
      • Extensions
    • iTerm 2
      • Theme
    • Fish
      • Plugins
    • Docker
      • Volumes
        • Mounting
    • Web Bundlers
      • Webpack
  • DevOps
    • Databases
      • Estimating Connections
    • Kubernetes
      • Terminology
      • kubectl Cheatsheet
      • Best Practices
      • Application Health Checks
      • Upgrades
      • Troubleshooting
      • Ruby on Rails
  • Awesome Products
  • Engineering Management
    • Overview
  • Software Architecture
    • Microservices
    • Event sourcing
    • Serverless
    • Centralized Authentication
  • Talks
    • Software Architecture
      • An Insider's Look at the Technology That Powers Shopify
      • Building Extensible Platforms
  • Documentation
  • Machine Learning
    • Terminology
    • Regression
    • Overfitting
  • Programming Languages
    • Constructs
    • Go
      • Syntax
    • Ruby
      • Ruby on Rails
        • Setup
        • Gotchas
        • Helpers
        • Libraries
        • Routing
        • Status Code Symbols
      • Debugging
        • Byebug
          • Cheatsheet
      • Libraries
      • Lazy Enumerators
      • Snippets
  • Version Control
    • Git
      • Conventions
        • Conventional Commits
        • gitmoji
  • Education
    • Programming & Computer Science
      • Courses
    • Design
      • Courses and Books
    • Frontend Development
  • HTTP
    • Status Codes
  • Design
    • Icons
      • Icon Sets
  • Arabic Content | محتوى عربي
    • Learning
      • مصادر لتعلم البرمجة و علوم الحاسب
    • Advice
      • إدارة المطورين: نبذة
      • نصائح تصحيح معتقدات خاطئة أمنيات للمبرمجين الجدد
  • Biology
    • Species
      • Ants
Powered by GitBook
On this page
  1. Machine Learning

Overfitting

Overfitting is when a machine learning model is fitting a data set too exactly and will cause it to fail when new data is introduced to it. In this case, the model fails (or simply has a high error rate) to deal with anything outside of the data set it was trained on.

tl;dr It's as if the model memorized the data set and knows nothing else outside of it.

Possible solutions:

Split the data set into one used for training, and one used for validation. If the model generalizes well, it should have similar loss metrics across the training and validation data sets.

PreviousRegressionNextProgramming Languages

Last updated 6 years ago