feature store
In plain wordsA shared library of ready-made input signals for models — such as a customer's spending last week — computed one consistent way and supplied both when models are trained and when they run live.
A data system that manages the engineered featuresfeature engineering, n.: Turning raw data into the inputs a model learns from, using knowledge of the problem; for example, converting a timestamp into 'hour of day' and 'is it a weekend'. used by machine-learning models, defining each feature once and supplying it consistently for training and for serving. It typically comprises a registry of feature definitions with owners and documentation; an offline store of historical values, from which training sets are built with point-in-time-correct joins that never use information from after the moment being predicted; an online store, a low-latency key–value database holding the latest values for real-time inferenceinference, n.: Drawing a conclusion from evidence. In today's AI it most often means running an already-trained model to get an answer, as opposed to training it.; and pipelines that compute features in batch or from event streams and keep both stores in sync. Its main purposes are preventing training–serving skewtraining–serving skew, n.: A mismatch between how data is prepared when a model is trained and when it is used for real, which makes the live model quietly worse than its tests promised. and data leakagedata leakage, n.: A mistake in building a model that lets it peek at clues it will not have in real use, such as hints about the very answer it must predict, so it shines in testing and then disappoints. and letting teams reuse features instead of recomputing them. Open-source and commercial examples include Feast, Tecton, Hopsworks and the feature stores of major cloud platforms.
Register the seven-day spend feature in the feature store so the fraud and credit teams use the same definition.
Origin
Popularized by Uber's 2017 description of its Michelangelo machine-learning platform, which included a shared feature store. The open-source Feast (Gojek with Google Cloud, 2019) and companies such as Tecton (founded by Michelangelo's creators) and Hopsworks made it a product category.
Usage note
Mostly used for tabular, event-based features in fraud detection, recommendation and pricing. Applications built on large language modelslarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code. rely more on vector databasesvector database, n.: A database built to store embeddings — lists of numbers that represent meaning — and to quickly find the stored items most similar to a new one. It is the memory behind many AI search tools. and document stores, though the two are sometimes combined.
Related
- feature engineeringfeature engineering, n.: Turning raw data into the inputs a model learns from, using knowledge of the problem; for example, converting a timestamp into 'hour of day' and 'is it a weekend'.
- training–serving skewtraining–serving skew, n.: A mismatch between how data is prepared when a model is trained and when it is used for real, which makes the live model quietly worse than its tests promised.
- MLOpsMLOps, n.: The working habits and tools that keep machine-learning models dependable inside real products once the experiment is over: automated testing, release, monitoring and retraining; the factory and maintenance crew around a model, not the model itself.
- data pipelinedata pipeline, n.: The chain of automated steps that collects raw data, cleans and transforms it, and delivers it in a usable form — the factory line that turns messy inputs into training material.
- machine learning pipelinemachine learning pipeline, n.: An automated chain of steps that takes raw data, prepares it, trains and tests a model and hands it over for use, so the whole process can be rerun the same way.
- vector databasevector database, n.: A database built to store embeddings — lists of numbers that represent meaning — and to quickly find the stored items most similar to a new one. It is the memory behind many AI search tools.
Sources
- Hermann, J. & Del Balso, M. (2017). Meet Michelangelo: Uber's Machine Learning Platform. Uber Engineering blog.
Mentioned in
feature engineeringfeature engineering, n.: Turning raw data into the inputs a model learns from, using knowledge of the problem; for example, converting a timestamp into 'hour of day' and 'is it a weekend'., machine learning pipelinemachine learning pipeline, n.: An automated chain of steps that takes raw data, prepares it, trains and tests a model and hands it over for use, so the whole process can be rerun the same way., MLOpsMLOps, n.: The working habits and tools that keep machine-learning models dependable inside real products once the experiment is over: automated testing, release, monitoring and retraining; the factory and maintenance crew around a model, not the model itself., machine learning technical debtmachine learning technical debt, n.: The hidden future cost of quick shortcuts in machine-learning systems — tangled data dependencies, messy glue code, unchecked feedback loops — that later make them hard and risky to change., training–serving skewtraining–serving skew, n.: A mismatch between how data is prepared when a model is trained and when it is used for real, which makes the live model quietly worse than its tests promised.