Diagnosing a New Rich-Output Task
Case context: You are designing a deep learning system for a hospital that takes an X-ray image (input X) and generates a full diagnostic text report (output Y). Your team is debating whether this is just a standard classification problem.
Question: Based on the end-to-end machine translation example, diagnose what type of learning this medical system represents and justify your conclusion.
Sample answer: This represents rich-output learning. Just like machine translation outputs a full sentence in French rather than a single number, this system outputs a full diagnostic text report. With the right (X-ray, text report) labeled pairs, it can be trained end-to-end to produce an output that is significantly richer than a simple classification score.
Key points:
- Classifies task as rich-output learning
- Identifies the output is a sentence/text rather than a single number
- Relies on having correct (input, output) labeled pairs
Rubric: The student must classify the task as rich-output learning and justify it by comparing the text report output to the single-number outputs of simpler tasks, referencing the translation example.
0
1
Tags
Data Science
Foundations of Large Language Models Course
Computing Sciences
D2L
Dive into Deep Learning @ D2L
Machine Learning
Deep Learning
Supervised Learning
Machine Learning Strategy
Machine Learning Yearning @ DeepLearning.AI
Related
In Ng's machine translation example of rich-output learning, what serves as the INPUT to the end-to-end system?
End-to-end machine translation trains directly on (English, French) labeled pairs without requiring hand-crafted intermediate representations.
Machine translation is a rich-output learning problem because the output is a _____ rather than a single number.
Match each component of Ng's machine translation example to its correct description.
Order the reasoning steps for deciding whether a new task qualifies for end-to-end rich-output learning.
Which statement best captures the 'accelerating trend in deep learning' Ng describes using machine translation as an example?
According to ML Yearning, 'rich outputs' in end-to-end deep learning are restricted to text sentences only.
Ng states that with the right (input, output) _____ pairs, you can sometimes learn end-to-end even when the output is rich.
Match each output type to its correct classification as a 'rich output' or 'simple output' in Ng's framework.
Order the stages of an end-to-end machine translation system as described in Ng's rich-output learning framework.
Explaining the Rich-Output Trend via Translation
Diagnosing a New Rich-Output Task
Defining Rich Outputs in Deep Learning