pre-trained model
[ˈpriː ˈtreɪnd ˈmɒdəl]
nounpl: pre-trained models
modelo pré-treinado
1. A machine learning model that has been previously trained on a large dataset and can be reused for new tasks with minimal additional training
The developer used a pre-trained model to classify images without training from scratch.
O desenvolvedor usou um modelo pré-treinado para classificar imagens sem treinar do zero.
2. An artificial neural network that has learned patterns from extensive data and serves as a starting point for transfer learning
BERT is a popular pre-trained model used for natural language processing tasks.
BERT é um modelo pré-treinado popular usado para tarefas de processamento de linguagem natural.
3. A computational architecture with weights and parameters already adjusted through prior training on relevant datasets
Fine-tuning a pre-trained model reduces the computational resources needed for new applications.
Ajustar um modelo pré-treinado reduz os recursos computacionais necessários para novas aplicações.
This is highly specialized terminology prevalent in AI/ML communities in both Brazil and the USA. In Brazil, as the tech sector grows, particularly in São Paulo, pre-trained models have become essential in startups and research institutions. The term is used identically in both English and Portuguese technical documentation, reflecting the international nature of AI development. American tech companies like OpenAI, Google, and Meta have popularized pre-trained models globally, making this term standard across all regions.
Related Idioms & Phrases
standing on the shoulders of giants (using existing pre-trained models)
don't reinvent the wheel (use a pre-trained model instead of training from scratch)
Look up more words on Fala2Me
The free English-Portuguese dictionary with real Brazilian accents, NYC slang, conjugator and more
Open Fala2Me →