7 papers
Theseus in the Graph: Towards Traceable Multi-Hop Graph Navigation
Eduin E. Hernandez, Luis F. Garcia, Nurassyl Askar +2
Multi-Hop Knowledge Graph Question Answering (KGQA) tasks require models to assemble relational evidence along paths in a KG to answer natural-language questions. However, existing…
How to Attain Communication-Efficient DNN Training? Convert, Compress, Correct
Zhong-Jing Chen, Eduin E. Hernandez, Yu-Chih Huang +1
This paper introduces CO3 -- an algorithm for communication-efficient federated Deep Neural Network (DNN) training. CO3 takes its name from three processing applied which reduce th…
Convert, compress, correct: Three steps toward communication-efficient DNN training
Zhong-Jing Chen, Eduin E. Hernandez, Yu-Chih Huang +1
In this paper, we introduce a novel algorithm, , for communication-efficiency distributed Deep Neural Network (DNN) training. is a joint training/com…
DNN gradient lossless compression: Can GenNorm be the answer?
Zhong-Jing Chen, Eduin E. Hernandez, Yu-Chih Huang +1
In this paper, the problem of optimal gradient lossless compression in Deep Neural Network (DNN) training is considered. Gradient compression is relevant in many distributed DNN tr…
Speeding-Up Back-Propagation in DNN: Approximate Outer Product with Memory
Eduin E. Hernandez, Stefano Rini, Tolga M. Duman
In this paper, an algorithm for approximate evaluation of back-propagation in DNN training is considered, which we term Approximate Outer Product Gradient Descent with Memory (Mem-…
Straggler Mitigation through Unequal Error Protection for Distributed Approximate Matrix Multiplication
Busra Tegin, Eduin. E. Hernandez, Stefano Rini +1
Large-scale machine learning and data mining methods routinely distribute computations across multiple agents to parallelize processing. The time required for the computations at t…
Straggler Mitigation through Unequal Error Protection for Distributed Matrix Multiplication
Busra Tegin, Eduin E. Hernandez, Stefano Rini +1
Large-scale machine learning and data mining methods routinely distribute computations across multiple agents to parallelize processing. The time required for computation at the ag…