C2W1 Quiz bad answers

I got notice to my answer that more data won’t reduce the bias. I think, it is wrong!

Q.: A model developed for a project is presenting high bias. One of the sponsors of the project offers some resources that might help reduce the bias. Which of the following additional resources has a better chance to help reduce the bias?
my_A.: Gather more data for the project.

Collecting more data for a project is more likely to reduce the load (bias) on the model. Collecting more data can help create a more diverse and representative dataset, allowing the model to better match actual data and reduce load. Increasing computing power through the use of graphics processing units (GPUs) or increasing the number of data sources (the other options of an answer) can improve the model’s training performance, but will not necessarily reduce load.

Hi @Andrzej_Krynski

High bias is under fitting the data. If your algorithm is not able to learn from the data, how could adding more data is going to help solved the problem?

Yes, this is not the question of quantity but rather of quality. So gathering more, different data, by it’s quality, can lead to better complexity of the model.

Hi @Andrzej_Krynski ,

Why would you train your model with poor quality data in the first place?

The quiz question here is reflecting back what Prof. Ng was talking about on Bias/variance and the Basic recipe for machine learning. If you refer to those lecture videos, you would see how he pointing out the different ways to solve these bias/variance problem.

1 Like