Research map / Probabilistic and scientific modeling / Molecules and proteins
ApproxIFER: A Model-Agnostic Approach to Resilient and Robust Prediction Serving Systems
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 rationally encoded queries and interpolation-based recovery instead of learned parity models | ApproxIFER versus ParM · MNIST, K=8 and K=12, S=1; ResNet-18 | accuracy improvement over ParM: 19%; 36% | Helped |
| Use rationally encoded queries and interpolation-based recovery instead of learned parity models | ApproxIFER versus ParM · Fashion-MNIST, K=8 and K=12, S=1; ResNet-18 | accuracy improvement over ParM: 7%; 17% | Helped |
| Use rationally encoded queries and interpolation-based recovery instead of learned parity models | ApproxIFER versus ParM · CIFAR-10, K=8 and K=12, S=1; ResNet-18 | accuracy improvement over ParM: 51%; 58% | Helped |
| Tolerate multiple stragglers using ApproxIFER's rational interpolation decoding | ApproxIFER · MNIST, S=2,3; ResNet-18 | accuracy loss versus best case: not more than 9.4% | Hurt |
| Tolerate multiple stragglers using ApproxIFER's rational interpolation decoding | ApproxIFER · Fashion-MNIST, S=2,3; ResNet-18 | accuracy loss versus best case: 8% | Hurt |
| Tolerate multiple stragglers using ApproxIFER's rational interpolation decoding | ApproxIFER · CIFAR-10, S=2,3; ResNet-18 | accuracy loss versus best case: 4.4% | Hurt |
| Recover predictions despite Byzantine workers by locating and excluding erroneous outputs | ApproxIFER · MNIST, K=12, up to E=3; ResNet-18 | accuracy loss versus best case: not more than 6% | Hurt |
| Recover predictions despite Byzantine workers by locating and excluding erroneous outputs | ApproxIFER · Fashion-MNIST, K=12, up to E=3; ResNet-18 | accuracy loss versus best case: 4% | Hurt |
| Recover predictions despite Byzantine workers by locating and excluding erroneous outputs | ApproxIFER · CIFAR-10, K=12, up to E=3; ResNet-18 | accuracy loss versus best case: 4.2% | Hurt |
| Use ApproxIFER with E=2 Byzantine adversaries across network architectures | ApproxIFER · CIFAR-10, VGG-16, ResNet-34, ResNet-50, DenseNet-161, GoogLeNet | accuracy loss versus best case: not more than 5% | Hurt |
Closest papers by meaning
- Weight for Robustness: A Comprehensive Approach towards Optimal Fault-Tolerant Asynchronous ML (NeurIPS 2024 · similarity 0.84)
- Coded Computing for Resilient Distributed Computing: A Learning-Theoretic Framework (NeurIPS 2024 · similarity 0.84)
- From Inexact Gradients to Byzantine Robustness: Acceleration and Optimization under Similarity (NeurIPS 2026 · similarity 0.84)
- Fault Tolerant ML: Efficient Meta-Aggregation and Synchronous Training (ICML 2024 · similarity 0.83)
- Dynamic Byzantine-Robust Learning: Adapting to Switching Byzantine Workers (ICML 2024 · similarity 0.83)
- Byzantine Robustness and Partial Participation Can Be Achieved at Once: Just Clip Gradient Differences (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.