Inference vs Prediction

Andrew introduces the idea of inference in MLS 2. However, I don’t understand this because as far as I can tell, we’re still doing prediction. I “googled” inference vs prediction, and inference seems to mean understanding “the relationship between the predictor variables and the response variable in an existing dataset.” At least as of midway through Week 1, we don’t seem to be doing this. Can someone explain? Thanks!

Sorry. I should have waited until I had watched the next video because I think I was able to determine the difference myself! I think Andrew uses inference not in an “inference vs prediction” way but to mean how the output of a layer is inferred from its input. Right?

Hi @xilan,

Let’s take a step back from the video and look at dictionaries’ definitions for the term inference.

Screenshot from 2022-07-10 12-17-05

In our case, the following terms used in the definitions should refer to the data: “information”, “known facts”, “evidence”.

As for the terms “guess”, “opinion” and “conclusion”, this is about what we want from the data, and if that is to make “prediction”, then in my opinion, “to infer” means “to predict”.