Hierarchical clustering can be broadly classified into two types based on how the hierarchy is constructed. What are these two types?
- Agglomerative and Divisive
- Linear and Non-linear
- QuickSort and MergeSort
- Recursive and Iterative
Hierarchical clustering can be Agglomerative (bottom-up) or Divisive (top-down) based on how clusters are merged or divided in the hierarchy.
You're analyzing data from a shopping mall's customer behavior and notice that there are overlapping clusters representing different shopping patterns. To model this scenario, which algorithm would be most suitable?
- K-Means Clustering
- Decision Trees
- Breadth-First Search
- Radix Sort
K-Means Clustering is commonly used for clustering tasks, such as identifying distinct shopping patterns. It groups data into clusters based on similarity, making it suitable for analyzing customer behavior data with overlapping patterns.
A company wants to determine the best version of their website homepage among five different designs. They decide to show each version to a subset of visitors and observe which version results in the highest user engagement. This problem is analogous to which classical problem in reinforcement learning?
- Multi-Armed Bandit
- Q-Learning
- Deep Q-Network (DQN)
- Policy Gradient Methods
This scenario is analogous to the Multi-Armed Bandit problem, where a decision-maker must choose between multiple options to maximize cumulative reward, akin to selecting the best website version for maximum user engagement.
Consider a robot that learns to navigate a maze. Instead of learning the value of each state or action, it tries to optimize its actions based on direct feedback. This approach is most similar to which reinforcement learning method?
- Monte Carlo Methods
- Temporal Difference Learning (TD)
- Actor-Critic Method
- Q-Learning
In this context, the robot is optimizing actions based on direct feedback, which is a characteristic of the Actor-Critic method. This method combines value-based and policy-based approaches, making it similar to the situation described.
Dimensionality reduction techniques, like PCA and t-SNE, are essential when dealing with the ________ curse.
- Overfitting
- Bias-Variance Tradeoff
- Curse of Dimensionality
- Bias
The "Curse of Dimensionality" refers to the increased complexity and sparsity of data in high-dimensional spaces. Dimensionality reduction techniques, such as PCA (Principal Component Analysis) and t-SNE, are crucial to mitigate the adverse effects of this curse.
A robot is navigating a maze. Initially, it often runs into walls or dead-ends, but over time it starts finding the exit more frequently. To achieve this, the robot likely emphasized ________ in the beginning and shifted towards ________ over time.
- Exploration, Exploitation
- Breadth-First Search
- Depth-First Search
- A* Search
In the context of reinforcement learning, the robot employs "exploration" initially to discover the maze, and as it learns, it shifts towards "exploitation" to choose actions that yield higher rewards, like finding the exit.
In reinforcement learning, the agent learns a policy which maps states to ________.
- Actions
- Rewards
- Values
- Policies
In reinforcement learning, the agent learns a policy that maps states to optimal actions, hence filling in the blank with "Policies" is accurate. This policy helps the agent make decisions in various states.
You are working on a dataset with a large number of features. While some of them seem relevant, many appear to be redundant or irrelevant. What technique would you employ to enhance model performance and interpretability?
- Data Normalization
- Feature Scaling
- Principal Component Analysis (PCA)
- Recursive Feature Elimination (RFE)
Principal Component Analysis (PCA) is a dimensionality reduction technique that can help reduce the number of features while preserving the most important information. It enhances model performance by eliminating redundant features and improves interpretability by transforming the data into a new set of uncorrelated variables.
Which term describes a model that has been trained too closely to the training data and may not perform well on new, unseen data?
- Bias
- Generalization
- Overfitting
- Underfitting
Overfitting is a common issue in machine learning where a model becomes too specialized to the training data and fails to generalize well to new data. It's essential to strike a balance between fitting the training data and generalizing to unseen data.
While LSTMs have three gates, the GRU simplifies the model by using only ________ gates.
- 1
- 2
- 3
- 4
Gated Recurrent Units (GRUs) simplify the model by using only two gates: an update gate and a reset gate, as opposed to the three gates in LSTMs.