Coded Sequential Matrix Multiplication For Straggler Mitigation

Authors: Nikhil Krishnan Muralee Krishnan, Seyederfan Hosseini, Ashish Khisti

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental These are further validated by experiments, where we implement our schemes to train neural networks.
Researcher Affiliation Academia M. Nikhil Krishnan University of Toronto nikhilkrishnan.m@gmail.com Erfan Hosseini University of Toronto ehosseini2108@gmail.com Ashish Khisti University of Toronto akhisti@ece.utoronto.ca
Pseudocode Yes Algorithm 1: Algorithm used by master to assign mini-tasks in the DIP coding scheme
Open Source Code No The paper does not provide an explicit statement or link for the open-sourcing of its methodology's code.
Open Datasets Yes Training is performed for 250 iterations over MNIST dataset with a batch size of 1024 using SGD.
Dataset Splits No The paper mentions training and testing but does not explicitly provide details about validation dataset splits.
Hardware Specification Yes We use four virtual machines with 8GB of RAM and 4 v CPUs as workers and one more machine with 16GB of RAM and 8 v CPUs as the master.
Software Dependencies No The paper mentions 'mpi4py [19] and Num Py' but does not specify their version numbers.
Experiment Setup Yes Each NN model is fully connected with two hidden layers of size 3000 followed by a Re LU activation. Training is performed for 250 iterations over MNIST dataset with a batch size of 1024 using SGD.