Can experts predict the best ML approach in advance?
Question: For a new machine learning problem, how well can an experienced practitioner usually predict the best approach before trying it?
Sample answer: Even a very experienced practitioner usually cannot predict the best solution with high accuracy at the start. In practice, they often need to test many different ideas and compare results before finding a strong approach.
Key points:
- Prior experience does not make the best choice obvious in advance.
- Several ideas usually need to be tried.
- Progress comes from experimentation and comparison.
Rubric: The answer should say that experts still cannot reliably know the best approach in advance and typically need to experiment with many alternatives before succeeding.
0
1
Tags
D2L
Dive into Deep Learning @ D2L
Machine Learning
Deep Learning
Supervised Learning
Data Science
Machine Learning Strategy
Machine Learning Yearning @ DeepLearning.AI
Related
Why do ML teams usually test several approaches on a new problem?
The Best ML Approach Is Usually Obvious Before Testing
For a brand-new machine learning task, it is usually _____ to know beforehand which method will work best.
Match Each Situation to the Correct Machine Learning Idea
Order the stages of a typical iterative machine learning workflow when the best solution is not obvious.
How many candidate ideas does an experienced ML practitioner often test before finding a good solution?
Choosing the best ML approach before any experiments is easy for experienced teams and mainly difficult for beginners.
What do teams often test many of before finding a workable machine-learning approach?
Match each development claim to its implication.
Put the reasoning in order for why ML work usually proceeds by iteration.
Why Machine Learning Requires Repeated Experimentation
Choosing an ML strategy for a new prediction task.
Can experts predict the best ML approach in advance?