Risk of the Learned Hypothesis
Compare notions of learnability by asking what each definition is useful for.
Start with the Goal
The word learnability does not become useful merely because someone has defined it. A definition is useful when it helps achieve a particular goal. That is the central discipline for this topic: state the goal first, then decide whether a definition supports it.
Three Definitions as Three Comparisons
Suppose three definitions of learnability are being considered. The source material does not provide their mathematical details, so the important task here is not to reconstruct them. Instead, compare what each definition is useful for. Ask what the definition directs your attention toward, what question it lets you ask, and whether that question matches your goal.
| Definition under consideration | Question to ask | Evidence of usefulness |
|---|---|---|
| Definition 1 | What does this definition allow us to compare for our stated goal? | The comparison helps address the goal |
| Definition 2 | What does this definition make visible that another definition may not? | The resulting information is relevant to the goal |
| Definition 3 | What decision or judgment would this definition support? | The definition supports the intended purpose |
A goal-based framework for comparing three definitions. The entries are questions for analysis, not mathematical descriptions of the definitions.
The three definitions should not be ranked in the abstract. A definition may be useful for one purpose and less useful for another. The relevant judgment is always connected to the goal.
Focusing on the Learned Hypothesis
To focus on the risk of the learned hypothesis is to treat the learned hypothesis as the central object of the comparison. Rather than discussing learnability only as an abstract definition, ask what the chosen notion tells you about the learned hypothesis and whether that information serves the stated learning goal.
This focus does not by itself tell you that one definition is superior. It tells you what to place at the center of the inquiry: the learned hypothesis and the purpose for which you want to understand it.
A Goal-Based Worked Example
Choosing a comparison question
You are asked to compare three definitions of learnability, but no specific learning goal has been stated. What should you do before deciding which definition is most useful?
Identify the missing information: The missing information is the purpose of the comparison. Without a stated goal, there is no responsible basis for calling one definition more useful than another.
State a possible goal: For this generated example, choose the goal of understanding what a definition tells us about the learned hypothesis.
Choose the comparison question: Ask: What does each definition help us say about the learned hypothesis for this goal?
Compare the definitions: Examine each definition using the same goal-based question. Do not add a universal ranking that the comparison does not justify.
The suitable comparison question depends on the stated goal. If the goal is to understand the learned hypothesis, evaluate each definition by how useful it is for that purpose.
The example does not claim that one of the three definitions wins. Its lesson is procedural: make the goal explicit, formulate a comparison question that serves that goal, and then judge usefulness in that context.
Questions That Keep the Comparison Honest
- What is the learning goal?
- What should the definition help us understand or decide?
- What is the learned hypothesis doing in this comparison?
- Which comparison question directly serves the goal?
- What information would count as useful for that purpose?
- Am I evaluating the definitions in relation to the goal, or assuming a universal ranking?
Mistakes in Comparing Learnability
Choosing a definition before stating the goal
Usefulness depends on the purpose the definition is intended to support.
Fix:
State the learning goal first, then ask whether the definition helps achieve it.Assuming one definition is always superior
The source material explicitly warns against assuming that one definition is best for every purpose.
Fix:
Evaluate each definition in relation to the particular goal.Treating the risk of the learned hypothesis as identical to the general comparison of learnability
The two topics are connected, but the source material distinguishes them.
Fix:
Keep the learned hypothesis as the focus when appropriate, while still asking what the broader definition is useful for.Inventing mathematical details that have not been supplied
This section deliberately does not provide mathematical details; its purpose is to practice goal-based comparison.
Fix:
Compare the definitions using their stated material in the actual discussion, rather than filling gaps with unsupported details.
Practice the Decision
A discussion presents three definitions of learnability and asks which one is useful. The only stated goal is to understand the learned hypothesis. Write one comparison question that fits this goal, and explain why a universal ranking would be inappropriate.
Hints
- Begin your question with what each definition helps you understand.
- Mention the learned hypothesis explicitly.
- Explain usefulness in relation to the stated goal rather than in isolation.
What do you think happens?
If the goal changes, should your judgment about which definition is useful automatically remain unchanged?
Reveal answer
Answer: No, usefulness must be reconsidered in relation to the new goal.
The central principle is to judge a definition by the purpose it is intended to support, not by an assumed universal ranking.
Key Takeaways
- A definition of learnability is useful when it helps achieve a particular goal.
- Always state the goal before deciding whether a definition is useful.
- Compare the three definitions by asking what each one contributes to the stated purpose.
- Do not assume that one definition is best for every purpose.
- Focusing on the risk of the learned hypothesis means placing the learned hypothesis at the center of a goal-based comparison while keeping that topic distinct from the broader comparison of learnability.
Key Takeaways
- State the learning goal before comparing definitions of learnability.
- Judge usefulness in relation to the goal, not through a universal ranking.
- Use a comparison question that makes clear what each definition is meant to help you understand.
- Treat the risk of the learned hypothesis as related to, but distinct from, the broader comparison of learnability.
- When the learned hypothesis is the focus, evaluate the comparison by whether it supports the stated purpose.