WorksheetsLearning Problems & Basics
Total questions: 50
Worksheet time: 25mins
A learning problem is defined by which three components?
Input, Output, Memory
Algorithm, Data, Result
Knowledge, Skill, Ability
Task, Performance, Experience
The task in a learning problem refers to:
Training data
What the system should learn to do
Accuracy measure
Output format
Performance measure evaluates:
Learning speed
Type of algorithm
Size of dataset
Success of learning
Training experience mainly consists of:
Heuristics
Goals
Examples or data
Rules
Learning from labeled data is called:
Unsupervised learning
Reinforcement learning
Supervised learning
Heuristic learning
Learning improves system performance by:
Experience
Memorization only
Hard coding
Guessing
A learning agent adapts based on:
Experience
Errors
Rules
Memory size
Which is NOT a learning problem component?
Compiler
Task
Experience
Performance
Email spam detection is an example of:
Learning
Planning
Reasoning
Searching
Machine learning mainly focuses on:
Programming rules
Hardware design
Manual coding
Learning from data
One key issue in learning systems is:
All of the above
Overfitting
Data scarcity
Data representation
One key issue in learning systems is:
All of the above
Overfitting
Data scarcity
Data representation
Overfitting occurs when a model:
Performs poorly on training data
Has fewer features
Performs well only on training data
Performs well on new data
Generalization means:
Applying learned knowledge to unseen data
Memorizing data
Removing features
Reducing data size
Bias in learning refers to:
Preference for certain hypotheses
Noise
Random choice
Error in data
Noise in data can cause:
No effect
Faster training
Overfitting
Perfect learning
Concept learning involves:
Learning algorithms
Learning rules or categories
Learning hardware
Learning functions
A concept can be represented as:
Set of objects
Boolean function
Rule
All of the above
Positive examples are:
Negative samples
Errors
Instances belonging to the concept
Incorrect data
Negative examples are:
Instances not belonging to the concept
Correct instances
Training data only
Output data
Concept learning is mainly used in:
Planning
Searching
Classification
Sorting
Version space is defined as:
All possible algorithms
Set of all hypotheses
Hypotheses consistent with training data
Final hypothesis
Candidate Elimination algorithm finds:
Single hypothesis
Optimal tree
Entire version space
Only general hypothesis
The most specific hypothesis is stored in:
G boundary
Training set
Version space
S boundary
The most general hypothesis is stored in:
S boundary
Memory
Dataset
G boundary
Candidate Elimination works with:
Only positive examples
Unlabeled data
Only negative examples
Both positive and negative examples
Version space is defined as:
All possible algorithms
Final hypothesis
Set of all hypotheses
Hypotheses consistent with training data
Candidate Elimination algorithm finds:
Single hypothesis
Only general hypothesis
Optimal tree
Entire version space
Candidate Elimination works with:
Unlabeled data
Only positive examples
Only negative examples
Both positive and negative examples
S boundary represents:
Most specific hypotheses
Most general hypotheses
Random hypotheses
Invalid hypotheses
G boundary represents:
Most specific hypotheses
Final hypothesis
Most general hypotheses
Training examples
Version space becomes empty when:
Dataset is large
Learning completes
Hypotheses conflict
Data is consistent
Main drawback of Candidate Elimination is:
Complexity
Sensitivity to noise
Slow speed
Memory usage
Inductive bias is:
Data error
Learning preference
Output error
Algorithm speed
Bias helps in:
Removing data
Increasing noise
Random guessing
Limiting hypothesis space
Decision tree bias prefers:
Deeper trees
Larger datasets
Smaller trees
Random trees
Inductive bias affects:
Hardware
Learning outcome
Output format
Input size
Decision tree bias prefers: A) Deeper trees B) Smaller trees C) Random trees D) Larger datasets
Larger datasets
Smaller trees
Random trees
Deeper trees
Inductive bias affects: A) Learning outcome B) Input size C) Output format D) Hardware
Hardware
Output format
Input size
Learning outcome
A decision tree is used for: A) Sorting B) Classification C) Searching D) Scheduling
Classification
Sorting
Searching
Scheduling
Internal nodes represent: A) Output labels B) Decisions/tests C) Data D) Rules
Data
Rules
Output labels
Decisions/tests
Leaf nodes represent: A) Attributes B) Conditions C) Class labels D) Algorithms
Algorithms
Conditions
Class labels
Attributes
ID3 algorithm uses: A) Entropy B) Information Gain C) Gini index D) Accuracy
Gini index
Information Gain
Entropy
Accuracy
Representation in learning means: A) Data storage B) Knowledge encoding C) Output format D) Algorithm selection
Knowledge encoding
Output format
Algorithm selection
Data storage
Heuristic is: A) Exact solution B) Rule of thumb C) Random guess D) Dataset
Exact solution
Rule of thumb
Random guess
Dataset
Heuristic search improves: A) Accuracy B) Speed of search C) Memory D) Data size
Speed of search
Memory
Accuracy
Data size
Best-first search uses: A) Random selection B) FIFO C) Heuristic evaluation D) Stack
FIFO
Stack
Heuristic evaluation
Random selection
A algorithm is:* A) Blind search B) Heuristic search C) Random search D) Exhaustive search
Heuristic search
Blind search
Exhaustive search
Random search
Heuristic function estimates: A) Exact cost B) Remaining cost C) Data size D) Accuracy
Accuracy
Remaining cost
Exact cost
Data size
Heuristic search is mainly used in: A) Optimization problems B) Sorting C) File handling D) Compilation
Sorting
File handling
Optimization problems
Compilation
