# one-shot learning

> object categorization problem that aims to learn information about object categories from one training sample

**Wikidata**: [Q7092335](https://www.wikidata.org/wiki/Q7092335)  
**Wikipedia**: [English](https://en.wikipedia.org/wiki/One-shot_learning_(computer_vision))  
**Source**: https://4ort.xyz/entity/one-shot-learning

## Summary
One-shot learning is an object categorization problem within machine learning that aims to learn information about object categories from only a single training sample. It is a specialized subclass of few-shot learning, designed to enable systems to recognize new objects or tasks without requiring the extensive datasets typically needed for traditional machine learning models.

## Key Facts
*   **Definition:** An object categorization problem focused on learning from one training sample.
*   **Parent Classifications:** Subclass of `machine learning` and `few-shot learning`.
*   **Instance Type:** Classified as a `field of study`.
*   **Wikipedia Scope:** Titled "One-shot learning (computer vision)" on English Wikipedia.
*   **Global Presence:** Covered in 5 sitelinks across languages including English, Arabic, Catalan, Farsi, and Serbian.
*   **Identifiers:** Freebase ID `/m/03hnhvz`; Microsoft Academic ID `2778034222` (discontinued).
*   **Also Known As:** "apprentissage par un seul exemple" (French), "ワンショット学習" (Japanese), and "aprendizado em um tiro" (Portuguese).

## FAQs
### Q: How does one-shot learning differ from traditional machine learning?
A: While traditional machine learning typically requires extensive datasets and explicit instructions to perform tasks, one-shot learning focuses on recognizing object categories or patterns from a single training sample.

### Q: What is the relationship between one-shot learning and few-shot learning?
A: One-shot learning is a specific subclass of few-shot learning. While few-shot learning covers learning from a small number of examples, one-shot learning is the extreme case where the system learns from exactly one sample.

### Q: What is the primary application of one-shot learning?
A: The primary application is object categorization, where the system must identify and classify objects after being exposed to just one training image or example.

## Why It Matters
One-shot learning represents a critical evolution in artificial intelligence, addressing the "data hunger" problem that plagues conventional machine learning models. In traditional machine learning, algorithms depend on massive datasets to achieve accuracy, a process that is often resource-intensive, time-consuming, or impossible in scenarios where data is scarce or rare (such as identifying rare diseases or specific industrial defects). By enabling systems to learn valid information from a single training sample, one-shot learning mimics human cognitive abilities more closely—humans can recognize a new object after seeing it just once.

This capability allows computer systems to perform tasks without explicit instructions for every scenario and significantly reduces the overhead of data collection and labeling. As a subclass of few-shot learning, it plays a pivotal role in making machine learning more adaptable and efficient, broadening the applicability of AI to fields where data is limited but rapid adaptation is required.

## Notable For
*   **Extreme Data Efficiency:** Distinguishes itself by requiring only one training sample, unlike standard deep learning models.
*   **Cognitive Similarity:** Notable for attempting to replicate the human ability to learn concepts instantly.
*   **Categorization Focus:** Specifically targets the problem of object categorization in computer vision.
*   **Broad Linguistic Reach:** Recognized as a distinct concept globally with specific aliases in multiple languages including French, Japanese, and Portuguese.

## Body

### Definition and Scope
One-shot learning is defined as an object categorization problem. Its primary objective is to learn information about object categories based on a single training sample. This distinguishes it from standard classification tasks which rely on hundreds or thousands of examples to train a model. It is categorized as a specific field of study within the broader discipline of computer science.

### Hierarchy and Classification
In the hierarchy of machine learning, one-shot learning sits as a specialized technique.
*   **Subclass of:** It is a direct subclass of `machine learning` (the study of algorithms that perform tasks without explicit instructions) and `few-shot learning` (learning from a few examples).
*   **Parent Relationship:** It falls under the class of `few-shot learning`, which serves as the broader approach for learning from limited data.

### Academic and Technical Identifiers
The concept is formally recognized in various knowledge bases and academic repositories:
*   **Wikipedia:** The English Wikipedia entry is titled "One-shot learning (computer vision)," indicating its strong association with visual recognition tasks.
*   **Wikidata:** The entity has a sitelink count of 5, connecting it to Wikipedia articles in Arabic (`ar`), Catalan (`ca`), English (`en`), Farsi (`fa`), and Serbian (`sr`).
*   **Freebase:** The entity is indexed under the identifier `/m/03hnhvz`.
*   **Microsoft Academic:** Historically indexed under ID `2778034222` (service now discontinued).