What is embedding in machine learning, and how does it work using ChromaDB?

Embedding in machine learning is used to transform multi-di­men­sion­al objects such as images, text, videos, or audio into vectors. This allows machine learning models to recognize and cat­e­go­rize them more ef­fec­tive­ly. This technique is par­tic­u­lar­ly suc­cess­ful in vector databases like ChromaDB, where it has already been applied with great success.

What is embedding in machine learning?

Embedding in machine learning is a technique that systems use to represent real-world objects in math­e­mat­i­cal form, making them un­der­stand­able for ar­ti­fi­cial in­tel­li­gence (AI). These em­bed­dings simplify the rep­re­sen­ta­tion of real objects while pre­serv­ing their features and re­la­tion­ships to other objects. The method is used to train machine learning models in iden­ti­fy­ing similar objects, which can include natural text, images, audio data, or videos. These objects are referred to as high-di­men­sion­al data, as they often contain complex details, such as the numerous pixel color values in an image.

Strictly speaking, AI em­bed­dings are vectors. In math­e­mat­ics, vectors are series of numbers that define a point in a di­men­sion­al space.

The core idea of em­bed­dings in machine learning is that a search algorithm within a vector database iden­ti­fies two vectors that are as close to each other as possible. The more complex these vectors are, the more accurate the result tends to be when two vectors are similar. For this reason, embedding in machine learning involves vec­tor­iz­ing and comparing as many factors or di­men­sions as possible. To achieve this, a model is trained using large and diverse datasets.

When is embedding used in machine learning?

Em­bed­dings are primarily used in machine learning for large language models. The method embeds not just a word, but also its context, allowing solutions like ChatGPT to analyze word sequences, sentences, or entire texts. Below are some ap­pli­ca­tion options for embedding in machine learning:

• Better searches and queries: Embedding in machine learning can be used to make searches and queries more precise, enabling more accurate outputs.
• Con­tex­tu­al­iza­tion: More precise answers can also be provided through ad­di­tion­al con­tex­tu­al in­for­ma­tion.
• Cus­tomiza­tion: Large language models can be specified and in­di­vid­u­al­ized using AI em­bed­dings. This enables precise tailoring to specific concepts or terms.
• In­te­gra­tion: Em­bed­dings can be used to integrate data from external sources, making datasets more extensive and het­ero­ge­neous.

How does embedding work? (Example: ChromaDB)

A vector database is the best solution. These databases not only store data ef­fi­cient­ly but also allow queries that return similar results rather than exact matches. One of the most popular open-source vector databases is ChromaDB. It stores em­bed­dings for machine learning along with metadata, allowing them to be used later by large language models (LLMs). This solution helps to better il­lus­trate how em­bed­dings work. In general, only the three steps presented below are required.

Step 1: Create a new col­lec­tion

In the first step, a col­lec­tion is created that resembles the tables stored in a re­la­tion­al database. These are then converted into em­bed­dings. ChromaDB uses the all-MiniLM-L6-v2 trans­for­ma­tion as the default for em­bed­dings, but this setting can be adjusted to use a different model. For example, if a spe­cial­ized col­lec­tion is needed, choosing another model can better address specific re­quire­ments, such as pro­cess­ing technical texts or images. The flex­i­bil­i­ty in model selection makes ChromaDB highly versatile, whether for text, audio, or image data.

Step 2: Add new documents

Next, you add text documents with metadata and a unique ID to the new col­lec­tion. Once the col­lec­tion contains the text, it is au­to­mat­i­cal­ly converted into em­bed­dings by ChromaDB. The metadata serves as ad­di­tion­al in­for­ma­tion to refine queries later, such as by filtering based on cat­e­gories or time­stamps. This struc­tur­ing allows for the efficient man­age­ment of large datasets and helps find relevant results more quickly.

Step 3: Retrieve the documents you are looking for

In the third step, you can query texts or em­bed­dings in ChromaDB. The output will return results that are similar to your query. It is also possible to retrieve the desired documents by entering the metadata. The results are sorted by sim­i­lar­i­ty, so the most relevant matches appear at the top. Ad­di­tion­al­ly, you can optimize the query by setting sim­i­lar­i­ty thresh­olds or applying ad­di­tion­al filters to further increase precision.