Coded-InvNet for Resilient Prediction Serving Systems
Authors: Tuan Dinh, Kangwook Lee
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experimental results show that Coded-Inv Net can outperform existing approaches, especially when the compute resource overhead is as low as 10%. |
| Researcher Affiliation | Academia | 1Department of Computer Sciences, University of Wisconsin Madison, Madison, USA 2Department of Electrical and Computer Engineering, University of Wisconsin-Madison, Madison, USA. |
| Pseudocode | No | The paper describes the system architecture and training procedures using text and diagrams, but it does not include formal pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide a specific link to open-source code or explicitly state that the code will be released. |
| Open Datasets | Yes | We focus on the image classification task on popular datasets: MNIST (Deng, 2012), Fashion MNIST (Xiao et al., 2017), and CIFAR10 (Krizhevsky & Hinton, 2009). |
| Dataset Splits | No | The paper mentions using 'train set' and 'test set' for encoder training but does not provide specific percentages or sample counts for train/validation/test splits for the main experiments. |
| Hardware Specification | Yes | Training the encoder may take up to 7.5 hours (Res Net-301-based architecture with 150 epochs on a 48-GB RTX8000 GPU). All of these are measured on a 12-GB NVIDIA TITAN Xp GPU, 128-GB of DRAM, and 40 Intel Xeon E5-2660 CPUs. All experiments are run on an Amazon AWS EC2 cluster with GPU instances (p2.xlarge with NVIDIA K80). |
| Software Dependencies | No | The paper mentions 'MPI4py (Dalcin et al., 2011)' but does not provide a specific version number for MPI4py or other key software components used in the experiments. |
| Experiment Setup | Yes | We train our i-Res Net classifier with Manifold Mixup (with mixup coefficient 1). We, then, train the encoder function using Least Square GAN loss (Mao et al., 2017) plus 100 times L1 loss. We fix the batch size as 256 images |