Ensemble
Combining several models’ predictions to get a result better than any one alone.
Think of it like
Like “ask the audience” — the crowd’s averaged guess usually beats any single person’s.
Example
A team runs three different models on the same input and takes the majority vote, cutting the error rate of each individual model.
How it actually works
Ensembling works because different models make different mistakes, so averaging or voting cancels uncorrelated errors. It reliably boosts accuracy and robustness, but at multiplied compute and latency — you’re running many models instead of one — which is why it’s common in competitions and less so in cost-sensitive production.
For product teams
A dependable accuracy bump you pay for in compute; great for offline scoring, pricey for live serving.
For engineers
Aggregating predictions (vote/average/stack) from diverse models to reduce variance; cost scales with members.
Related
- Mixture model — A weighted, learned relative.
- Sampling — Averaging many best guesses.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome