Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time.
Mitchell WortsmanGabriel IlharcoSamir Yitzhak GadreRebecca RoelofsRaphael Gontijo LopesAri S. MorcosHongseok NamkoongAli FarhadiYair CarmonSimon KornblithLudwig SchmidtPublished in: CoRR (2022)
Keyphrases
- probabilistic model
- multiple models
- statistical model
- neural network model
- goodness of fit
- prior knowledge
- modeling framework
- statistical models
- experimental data
- high accuracy
- computational models
- metamodel
- classification models
- hybrid model
- accurate models
- domain models
- inference process
- analytical model
- bayesian inference
- decision theoretic
- parametric models
- mathematical model
- computational model
- monte carlo simulation
- mathematical models
- objective function
- bayesian model
- learned models
- random fields
- prediction model
- factor graphs
- linear combination
- gibbs sampler