Learn Before
Why Scored Decoding Often Uses Approximate Search
A system can assign a score to each possible output and still be unable to search every option exactly, because the space of candidates may be enormous. In that situation, the system uses an approximate search method to find a strong candidate quickly. Such methods are practical, but they do not guarantee finding the absolute highest-scoring output.
0
1
Tags
Data Science
Machine Learning
Deep Learning
Supervised Learning
Dive into Deep Learning @ D2L
Machine Learning Strategy
Machine Learning Yearning @ DeepLearning.AI
Related
Why Scored Decoding Often Uses Approximate Search
Distance-Based Scoring Function
Semantic Matching Scoring Function
In a speech-to-text model, what does Score_audio(phrase) represent?
True or False: Assigning scores to candidates does not eliminate the need for a separate step that finds the highest-scoring choice.
Ranking by Score
Match each scoring-function element to its role in a route-planning example.
Put the score-and-search workflow in order.
Why separate scoring from search in an AI system?
Use the score comparison to identify the source of the error.
What are the two parts of a typical scoring-and-search design?
When is the optimization-check idea applicable?
True or False: A common machine learning pipeline uses an exact scoring model followed by an exact maximization procedure.
Learn After
Beam Search as an Approximate Search Method
Why is it usually impractical to score every possible output sequence and choose the best one exactly?
True or False: Approximate search methods used during scored inference are guaranteed to return the highest-scoring result.
Because checking every possible output is impractical, a scored inference system must use a(n) _____ method to locate a high-scoring result.
Why is exact search impractical in a scored sentence-generation system?
Approximate search methods in scored inference are guaranteed to return the candidate S with the highest value of Score_A(S).
Search target in scored decoding
Match each inference term to its role in a scored output system.
Order the steps a speech transcription system uses to convert a recorded message into text.
Meaning of a Conditional Score in Classification
If each of N positions can be filled by any one of 50,000 words, then the number of distinct sequences of length N is (50,000)^N.
Approximate Search and Score Maximization
Match each search challenge to the concept that best describes it.
Order the reasoning steps that explain why approximate search is needed in scored inference.
Why exhaustive search is impractical in scored inference
Find out why an approximate decoder missed a better-scoring output.
State the cost of approximate search in large-scale scoring problems.