Data analysis and visualization: Practice Questions — Data Analysis and Visualization (NVIDIA-Certified Associate: Generative AI LLM)
Practice Questions: Data Analysis and Visualization for NVIDIA-Certified Associate: Generative AI LLM These multiple-choice questions focus on data...
Practice Questions: Data Analysis and Visualization for NVIDIA-Certified Associate: Generative AI LLM
These multiple-choice questions focus on data analysis and visualization, a key topic area for the NVIDIA-Certified Associate: Generative AI LLM exam. Each question tests foundational knowledge relevant to data preprocessing, feature engineering, GPU-accelerated data manipulation, and preparing datasets for machine learning workflows.
Which of the following is the primary purpose of feature engineering in data preprocessing?
- A) To reduce the size of the dataset by removing samples
- B) To create new input variables that improve model performance
- C) To visualize data distributions and correlations
- D) To convert categorical data into image data
Correct answer: B
Explanation: Feature engineering involves creating or transforming variables to better represent the underlying problem, which can improve the predictive power of machine learning models.
When using GPU-accelerated data manipulation libraries like RAPIDS cuDF, what is a key advantage compared to traditional CPU-based processing?
- A) It automatically generates visualizations
- B) It enables faster processing of large datasets through parallel computation
- C) It eliminates the need for data cleaning
- D) It reduces the dataset size by compressing data
Correct answer: B
Explanation: GPU acceleration leverages parallelism to speed up data manipulation tasks, which is especially beneficial for large-scale datasets used in AI model training.
Which visualization technique is most suitable for understanding the distribution of a single continuous variable?
- A) Scatter plot
- B) Histogram
- C) Heatmap
- D) Box plot
Correct answer: B
Explanation: Histograms show the frequency distribution of a continuous variable, making it easier to observe patterns such as skewness or modality.
In preparing datasets for training large language models, why is data normalization important?
- A) It converts text data into numerical vectors
- B) It ensures consistent scale across features to improve model convergence
- C) It removes stop words from text
- D) It visualizes feature importance
Correct answer: B
Explanation: Normalization rescales features to a common scale, which helps optimization algorithms converge more efficiently during training.
Which of the following is a common step in data preprocessing for text data used in generative AI models?
- A) Tokenization
- B) Principal Component Analysis (PCA)
- C) Image augmentation
- D) Feature scaling
Correct answer: A
Explanation: Tokenization splits text into smaller units (tokens), which is essential for processing and encoding text data in language models.
What is the main benefit of using interactive visualizations in data analysis?
- A) They reduce the dataset size
- B) They allow dynamic exploration of data patterns and relationships
- C) They automatically preprocess data
- D) They replace the need for feature engineering
Correct answer: B
Explanation: Interactive visualizations enable users to manipulate views, filter data, and drill down into details, facilitating deeper insights.
Which GPU-accelerated library is commonly used for data manipulation in NVIDIA AI workflows?
- A) TensorFlow
- B) RAPIDS cuDF
- C) Matplotlib
- D) Scikit-learn
Correct answer: B
Explanation: RAPIDS cuDF is a GPU-accelerated dataframe library designed for fast data manipulation, complementing NVIDIA AI development.
What is the role of data visualization in the context of preparing datasets for large language models?
- A) To train the model directly
- B) To identify data quality issues and understand feature distributions
- C) To encode text into embeddings
- D) To perform hyperparameter tuning
Correct answer: B
Explanation: Visualization helps detect anomalies, missing values, or biases in data, which is critical before training generative AI models.