iT邦幫忙

第 11 屆 iThome 鐵人賽

DAY 23
0
Google Developers Machine Learning

ML Study Jam Journey系列 第 23

Day 23 Art and Science of Machine Learning (cont.)

  • 分享至 

  • xImage
  •  

Logistic Regression

  • Use Cross Entropy typically

  • Two problems

    1. Weights will be driven to -inf and +inf the longer we train
    2. Near the asymptotes, gradient is really small

Training

  • Gradients can vanish - Use ReLu instead of sigmoid/tanh
  • Gradients can explode - Batch Normalization
  • ReLu layers can die - Lower learning rates

上一篇
Day 22 Art and Science of Machine Learning (cont.)
下一篇
Day 24 Art and Science of Machine Learning (cont.)
系列文
ML Study Jam Journey30
圖片
  直播研討會
圖片
{{ item.channelVendor }} {{ item.webinarstarted }} |
{{ formatDate(item.duration) }}
直播中

尚未有邦友留言

立即登入留言