Lifted Proximal Operator Machines

Authors: Jia Li, Cong Fang, Zhouchen Lin4181-4188

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on MNIST and CIFAR-10 datasets testify to the advantages of LPOM. We implement LPOM on fully connected DNNs and test it on benchmark datasets, MNIST and CIFAR-10, and obtain satisfactory results.
Researcher Affiliation Academia Jia Li, Cong Fang, Zhouchen Lin Key Laboratory of Machine Perception (MOE), School of EECS, Peking University, P. R. China jiali.gm@gmail.com; fangcong@pku.edu.cn; zlin@pku.edu.cn
Pseudocode Yes Algorithm 1 Solving LPOM; Algorithm 2 Solving (32).
Open Source Code No The paper does not provide an explicit statement or link for open-sourcing the code for LPOM.
Open Datasets Yes Experiments on MNIST and CIFAR-10 datasets testify to the advantages of LPOM. For the MNIST dataset, we use 28 28 = 784 raw pixels as the inputs. It includes 60,000 training images and 10,000 test images. http://yann.lecun.com/exdb/mnist/
Dataset Splits No For the MNIST dataset, we use 28 28 = 784 raw pixels as the inputs. It includes 60,000 training images and 10,000 test images. The paper mentions training and testing, but no explicit validation split.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments.
Software Dependencies No We implement LPOM with MATLAB without optimizing the code. We use the SGD based solver in Caffe (Jia et al. 2014). No version numbers are provided for MATLAB or Caffe.
Experiment Setup Yes We run LPOM and SGD for 100 epochs with a fixed batch size 100. For LPOM, we simply set µi =20 in (18). For LPOM, we set µi = 100 in (18). For LPOM, we set µi = 20 for all the networks.