NVIDIA-Certified-Professional Accelerated Data Science Sample Questions:

1. You are analyzing a dataset containing millions of transaction records to detect fraudulent transactions. The dataset is highly imbalanced, with fraudulent cases making up only 0.5% of the data.
Which data analysis strategy will most effectively handle the imbalance while minimizing the risk of overfitting?

A) Random oversampling of the minority class
B) SMOTE (Synthetic Minority Over-sampling Technique)
C) Stratified random sampling
D) Random undersampling of the majority class

2. A data scientist is processing a dataset that is too large to fit into the memory of a single GPU. They decide to use Dask with cuDF to leverage multiple GPUs for accelerated computation.
Which of the following approaches ensures efficient parallelism when working with dask_cudf?

A) Assign explicit chunks of data to each worker by manually dividing the dataset before passing it to Dask.
B) Convert the dataset into a Dask cuDF dataframe and use persist() to keep it in GPU memory.
C) Use a single-threaded scheduler to avoid contention when running on multiple GPUs.
D) Use Dask futures to asynchronously submit GPU tasks and collect results when completed.

3. You are training a machine learning model using NVIDIA RAPIDS cuML and notice that the training process is significantly slower than expected. You suspect that there are bottlenecks in data movement and computation.
Which of the following techniques can best help you diagnose and resolve these bottlenecks?

A) Use cudf.DataFrame.to_pandas() to convert the dataset to a pandas DataFrame for analysis.
B) Use cuml.common.device_auto_mem_size() to check GPU memory usage and adjust batch sizes accordingly.
C) Move all data from GPU memory to CPU memory before training the model.
D) Reduce the number of features used in training without profiling the actual bottlenecks.

4. A data science team is deploying a deep learning model for real-time inference. The model is optimized for inference on an NVIDIA A100 GPU, but the team notices that inference latency is higher than expected.
Which of the following optimizations is most effective in reducing inference latency?

A) Increase the batch size significantly to improve GPU utilization.
B) Use mixed-precision inference with TensorRT to accelerate computation.
C) Enable CPU offloading to balance the workload between the CPU and GPU.
D) Reduce the model size by randomly pruning neurons without retraining.

5. You are working on a machine learning dataset with millions of rows and want to optimize performance on a GPU using CuDF. One of the features in your dataset represents categorical data with a limited number of unique values.
Which of the following data types should you use to maximize performance while reducing memory usage?

A) category
B) int64
C) bool
D) object (string)

Solutions: