Toy example of Logistic Regression with Tensorflow probability and the titanic dataset fails
- 阅读更多 关于 Toy example of Logistic Regression with Tensorflow probability and the titanic dataset fails
问题 I am learning tensorflow-probability and this is a toy example of logistic regression with the titanic dataset. My model does not seem to learn and the loss is nan. I don't understand why. Below you will find three different implementations, all return the same results. One uses a sigmoid activation, the second uses a DistributionLambda Layer with a Bernoulli distribution and the third a DistributionLambda Layer with a Beta distribution. Are there any corrections I should make to this code?