本文列举了在构建和训练机器学习模型过程中常见的十个错误做法，包括数据预处理不当、正则化缺失、批量大小设置不合理等问题，并提供了相应的解决建议。
没错的，又是传送门
大概分为以下几点：
1、You Forgot to Normalize Your Data
2、You Forgot to Check your Results
3、You Forgot to Preprocess Your Data
4、You Forgot to use any Regularization
5、You Used a too Large Batch Size
6、You Used an Incorrect Learning Rate
7、You Used the Wrong Activation Function on the Final Layer
8、Your Network contains Bad Gradients
9、You Initialized your Network Weights Incorrectly
10、You Used a Network that was too Deep
11、You Used the Wrong Number of Hidden Units
扫一扫
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
