Essay

Analyzing the Skepticism of End-to-End Autonomous Driving

Question: In the context of autonomous driving, explain why Andrew Ng is skeptical of using an end-to-end machine learning approach compared to a non-end-to-end approach. What specific factor drives this preference?

Sample answer: Andrew Ng is skeptical of end-to-end learning for autonomous driving primarily due to data availability constraints. He notes that the non-end-to-end architecture better matches the availability of data. Until significantly more end-to-end data becomes available, breaking the problem down using a non-end-to-end approach is considered significantly more promising.

Key points:

  • End-to-end approaches require massive amounts of end-to-end data.
  • A non-end-to-end architecture currently better matches the availability of data.
  • Until more end-to-end data is collected, the non-end-to-end approach is more promising.

Rubric: The response should identify data availability as the core issue and explain that the non-end-to-end architecture currently aligns better with the data that is available.

0

1

Updated 2026-05-27

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Machine Learning Yearning @ DeepLearning.AI