Research map / Learning theory and algorithms / Sample complexity
Least Squares Model Averaging for Distributed Data
Haili Zhang, Zhaobo Liu, Guohua Zou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| average over candidate models with Mallows weights instead of selecting one by AIC | simple aggregated estimator · Airline data (K=1000 subsets) | Mean MSPE (x10^-2): 6.263 -> 6.255 | Improved |
| use doubly simple aggregation (dMMA) instead of simple aggregation (MMA) | Mallows model averaging · Airline data (K=1000 subsets) | Mean MSPE (x10^-2): 6.255 -> 6.253 | Improved |
| use dMMA instead of MMA at large K | Mallows model averaging · Airline data (K=500 subsets) | Mean MSPE (x10^-2): 6.209 -> 6.206 | Improved |
| use dMMA instead of MMA at K=100 | Mallows model averaging · Airline data (K=100 subsets) | Mean MSPE (x10^-2): 6.159 -> 6.160 | Hurt |
| use dMMA instead of AIC for optimal rate | distributed model averaging · Airline data (K=1000 subsets) | Optimal rate: 0.260 -> 0.382 | Improved |
| increase number of subsets K from 1 to 100 | all nine methods · Airline data | Mean MSPE | Improved |
| increase number of subsets K from 100 to 1000 | all nine methods · Airline data | Mean MSPE: 6.159 -> 6.255 (MMA) | Hurt |
| use BIC-type selection or averaging instead of AIC/Mallows type | BIC, SBIC, dSBIC · Simulation (infinite order regression) | risk | Hurt |
| use model averaging instead of model selection | SAIC, SBIC, MMA vs AIC, BIC, Mallows · Simulation in-sample risk (n=50, 150) | normalized risk | Improved |
| use doubly simple aggregation instead of simple aggregation | model averaging estimators · Simulation out-of-sample risk | normalized risk | Improved |
| increase sample size n to 10000 | dMMA versus MMA · Simulation in-sample risk | normalized risk | No clear effect |
Closest papers by meaning
- Distributed Statistical Inference under Heterogeneity (JMLR 2023 · similarity 0.87)
- Model Averaging Is Asymptotically Better Than Model Selection For Prediction (JMLR 2022 · similarity 0.86)
- Debiased Distributed Learning for Sparse Partial Linear Models in High Dimensions (JMLR 2022 · similarity 0.86)
- Stability and L2-penalty in Model Averaging (JMLR 2024 · similarity 0.85)
- On the Necessity of Collaboration for Online Model Selection with Decentralized Data (NeurIPS 2024 · similarity 0.83)
- Distributed Least Squares in Small Space via Sketching and Bias Reduction (NeurIPS 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
