22 Questions
What is the primary goal of an agent in reinforcement learning?
To get the maximum positive rewards
What is the goal of supervised learning?
To learn the mapping function from input to output
What is a hypothesis in machine learning?
A candidate model that approximates a target function
What is hypothesis space?
A set of all possible legal hypothesis
What is version space?
A set of hypotheses that are consistent with the set of training examples
What is a most general hypothesis?
A hypothesis that covers none of the negative examples
What is the condition for a hypothesis to be consistent with the set of training examples?
h(x) = c(x) for all x ∈ D
What is the relationship between a hypothesis h and a more general hypothesis h'?
h is more general than h' if h'(x) = 1 then h(x) = 1
What is the definition of a most general hypothesis G?
The largest axis-aligned rectangle that covers all positive examples and no negative examples.
What is the main characteristic of a most specific hypothesis?
It covers none of the negative examples and there is no other hypothesis that covers no negative examples, such that the other hypothesis is more specific.
What is the Version Space of hypotheses?
The set of all hypotheses that are consistent with the training set.
What is the main purpose of the Vapnik-Chervonenkis dimension?
To guide the model selection process in machine learning.
What is the issue that arises when a model is too complex and fits the noise in the training data?
Overfitting
What is the main advantage of using a model with low VC dimension?
It is less prone to overfitting.
What is the purpose of model selection in machine learning?
To select the best inductive bias for learning
What happens when no hypothesis or multiple hypotheses are 1 for a given x?
The classifier rejects such cases
What is the advantage of the one-against-one strategy in multi-class classification?
It allows for the creation of multiple classifiers for each pair of classes
What is the purpose of inductive bias in machine learning?
To make assumptions to have a unique solution with the data
What is the problem with a highly complex model?
It is prone to overfitting
What is the goal of generalization in machine learning?
To perform well on unseen data
What is the problem with an ill-posed problem in machine learning?
There is no unique solution
What happens when the classifier assigns the same class to an instance in the one-against-one strategy?
The instance is assigned the majority class
Study Notes
Classification Strategies
- In the one-against-all strategy, we find K hypotheses (h1,..., hK) where only one of hi(x) is 1, and we assign the class Ci to x.
- If multiple hi(x) is 1, we say that the classifier rejects such cases.
One-Against-One (OAO) Strategy
- In OAO, a classifier is constructed for each pair of classes, resulting in K(K-1)/2 classifiers.
- An unknown instance is classified with the class getting the most votes, with ties broken arbitrarily.
Model Selection and Generalization
- Model selection refers to the process of picking a particular mathematical model from among different models.
- Inductive bias is the set of assumptions we make to have learning possible, as learning is an ill-posed problem.
- The primary goal of an agent in reinforcement learning is to improve performance by getting maximum positive rewards.
Supervised Learning
- Supervised learning involves using an algorithm to learn the mapping function from input variables (x) to an output variable (Y).
- The goal is to approximate the mapping function so well that the model can predict the output (y) for new input data (x).
Hypothesis and Hypothesis Space
- A hypothesis in machine learning is a candidate model that approximates a target function for mapping inputs to output.
- Hypothesis space is the set of all possible legal hypotheses.
Version Space
- Version Space consists of all hypotheses that are consistent with the set of training examples.
- A hypothesis is consistent with a set of training examples if it correctly classifies all the training examples.
Most General and Most Specific Hypothesis
- A most general hypothesis covers none of the negative examples and is the most generic hypothesis.
- A most specific hypothesis covers none of the negative examples and is the most specific hypothesis.
S, G, and the Version Space
- Any hypothesis h ∈ H between S (most specific hypothesis) and G (most general hypothesis) is a valid hypothesis with no errors and thus consistent with the training set.
- All such hypotheses h make up the Version Space of hypotheses.
Vapnik-Chervonenkis (VC) Dimension
- The Vapnik-Chervonenkis dimension is a model capacity measurement used in statistics and machine learning.
- It is an informal measure of a model’s capacity and is used to guide the model selection process.
Understand the concepts of one against all and one against one classification in machine learning, where a classifier is trained to distinguish one class from another.
Make Your Own Quizzes and Flashcards
Convert your notes into interactive study material.
Get started for free
