NVIDIA-Certified-Professional Accelerated Data Science Sample Questions:

1. A financial analyst is working with an irregularly spaced time-series dataset containing cryptocurrency transactions. The timestamps are not evenly distributed, with some periods having dense data and others having sparse entries. The analyst wants to visualize the data efficiently using GPU acceleration.
What is the best preprocessing approach before visualization?

A) Convert the dataset to Pandas and use df.resample() to aggregate by fixed time intervals.
B) Ignore the irregularity and plot the raw timestamps directly without any preprocessing.
C) Resample the time-series to a fixed frequency using cuDF.resample() and fill missing values.
D) Sort the data by timestamp and drop all sparse regions using df.dropna().

2. A data scientist is analyzing sales data for an e-commerce company that experiences strong seasonal trends (e.g., increased sales during holiday seasons). The goal is to accurately forecast future sales using GPU-accelerated data science techniques.
Which approach would be the most effective?

A) Compute a rolling average and manually adjust for seasonality based on previous peak sales months.
B) Use a Seasonal Autoregressive Integrated Moving Average (SARIMA) model and optimize it using NVIDIA RAPIDS.
C) Apply k-Means clustering to identify seasonal patterns and extrapolate future values.
D) Train a Decision Tree model using cuML to classify future sales trends.

3. A data scientist is working on a machine learning model for fraud detection. Due to the limited size of the dataset, they decide to generate synthetic data using NVIDIA RAPIDS AI and cuDF.
Which of the following approaches is the most efficient and effective for generating synthetic data while ensuring compatibility with RAPIDS AI workflows?

A) Train a generative adversarial network (GAN) using PyTorch and then use the generated samples in RAPIDS AI without any additional processing.
B) Use the RAPIDS cuML library to directly generate synthetic tabular data with controlled statistical properties.
C) Use numpy and pandas to generate synthetic data, then convert the DataFrame to cuDF before using it in RAPIDS AI workflows.
D) Use cuDF DataFrame operations to create new synthetic samples by applying random transformations (e.g., noise injection, permutation) to the existing dataset.

4. Which of the following scenarios are most appropriate for using GPU acceleration when working with large-scale datasets in machine learning? (Select two)

A) Training a large-scale natural language processing (NLP) model on text data with billions of words.
B) Running a large ensemble of simple models (e.g., random forests) on a dataset of 10 million rows.
C) Performing exploratory data analysis (EDA) on a dataset of 100,000 rows with 10 features.
D) Running a deep learning model for image classification with millions of labeled images.
E) Using decision trees for a classification task with a dataset of 1 million rows and 20 features.

5. Which tools or technologies from NVIDIA are essential for implementing an efficient MLOps pipeline in production environments? (Select two)

A) NVIDIA TensorRT for efficient model inference
B) NVIDIA Triton Inference Server for managing deployment and serving models
C) NVIDIA CUDA for model training in cloud environments
D) NVIDIA NGC for storing and sharing machine learning datasets
E) NVIDIA DLA (Deep Learning Accelerator) for model deployment

Solutions: