Selected article for: "initial learning rate and learning rate"

Author: Xuehai He; Xingyi Yang; Shanghang Zhang; Jinyu Zhao; Yichen Zhang; Eric Xing; Pengtao Xie
Title: Sample-Efficient Deep Learning for COVID-19 Diagnosis Based on CT Scans
  • Document date: 2020_4_17
  • ID: l3f469ht_58
    Snippet: 1) Implementation Details: For classifiers trained from scratch, the Adam [56] optimizer is used with an initial learning rate of 0.0001 and a mini-batch size of 16. The cosine annealing scheduler is applied on the optimizer with a period of 10 to adjust the learning rate across the training process. We train our models with 50 epochs. We initialize the weights with Kaiming Initialization [57] ......
    Document: 1) Implementation Details: For classifiers trained from scratch, the Adam [56] optimizer is used with an initial learning rate of 0.0001 and a mini-batch size of 16. The cosine annealing scheduler is applied on the optimizer with a period of 10 to adjust the learning rate across the training process. We train our models with 50 epochs. We initialize the weights with Kaiming Initialization [57] .

    Search related documents:
    Co phrase search for related documents
    • initial learning rate and model train: 1
    • learning rate and mini batch size: 1, 2
    • learning rate and model train: 1, 2, 3, 4, 5
    • learning rate and training process: 1, 2, 3, 4
    • mini batch size and training process: 1
    • model train and scratch train: 1
    • model train and training process: 1, 2, 3