Research map / Time series / Long-term forecasting
An Algorithmic Framework for the Optimization of Deep Neural Networks Architectures and Hyperparameters
Julie Keisler, El-Ghazali Talbi, Sandra Claudel, Gilles Cabriel
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 |
|---|---|---|---|
| use evolutionary search with joint mutation instead of random search (ablation) | DRAGON · M1 Monthly | MASE: 1.098 -> 1.073 | Improved |
| use evolutionary search with joint mutation instead of random search (ablation) | DRAGON · Tourism Monthly | MASE: 1.645 -> 1.450 | Improved |
| use evolutionary search with joint mutation instead of simulated annealing (ablation) | DRAGON · M1 Monthly | MASE: 1.141 -> 1.073 | Improved |
| use evolutionary search with joint mutation instead of simulated annealing (ablation) | DRAGON · Tourism Monthly | MASE: 2.640 -> 1.450 | Improved |
| joint instead of alternating architecture/hyperparameter mutation (ablation) | DRAGON · M1 Monthly | MASE: 1.080 -> 1.073 | No clear effect |
| joint instead of alternating architecture/hyperparameter mutation (ablation) | DRAGON · Tourism Monthly | MASE: 1.451 -> 1.450 | No clear effect |
| use random search instead of simulated annealing (ablation) | DRAGON · Tourism Monthly | MASE: 2.640 -> 1.645 | Improved |
| add the training seed as a searched hyperparameter (ablation) | DRAGON · M3 Quarterly | MASE: best MASE: 1.099 | Improved |
| add the training seed as a searched hyperparameter (ablation) | DRAGON · Electricity Weekly | MASE: best MASE: 0.652 | Improved |
| DRAGON versus AutoGluon-TS ensemble | DRAGON · Monash 27 datasets | datasets won (MASE): 13 wins for AutoGluon, 14 losses | Mixed |
| DRAGON versus AutoPytorch | DRAGON · Monash 27 datasets | datasets won (MASE): 7 wins for AutoPytorch, 20 losses | Improved |
Closest papers by meaning
- AutoGO: Automated Computation Graph Optimization for Neural Network Evolution (NeurIPS 2023 · similarity 0.85)
- Optimizing Memory Placement using Evolutionary Graph Reinforcement Learning (ICLR 2021 · similarity 0.85)
- Transferable Graph Optimizers for ML Compilers (NeurIPS 2020 · similarity 0.85)
- PACE: A Parallelizable Computation Encoder for Directed Acyclic Graphs (ICML 2022 · similarity 0.85)
- Accelerating neural network training: An analysis of the AlgoPerf competition (ICLR 2025 · similarity 0.84)
- How Graph Neural Networks Learn: Lessons from Training Dynamics (ICML 2024 · similarity 0.84)
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.
