Microsoft DP-100 - Designing and Implementing a Data Science Solution on Azure Certification Exam

Question #1 (Topic: demo questions)

DRAG DROP You need to define a process for penalty event detection. Which three actions should you perform in sequence? To answer, move the appropriate actions from the list of actionsto the answer area and arrange them in the correct order.

A.

Actions (Available Options)
Standardize to mono audio clips.
Vary the length of sliding windows between modeling epochs.
Vary the length of frequency bands between modeling epochs.
Use an Inverse Fourier transform on frequency changes over time.
Use a Fast Fourier transform on frequency changes over time.
Standardize to stereo audio clips.
Answer Area (Ordered Steps)
Step Sequence Action to Perform
Step 1 Vary the length of frequency bands between modeling epochs.
Step 2 Standardize to mono audio clips.
Step 3 Use an Inverse Fourier transform on frequency changes over time.

Correct Answer: A

Explanation:

Step Action Technical Significance
Step 1 Vary the length of frequency bands between modeling epochs. Adjusting or varying the configuration parameters of your frequency data distribution across epochs ensures robust data diversity and structural feature balancing before deeper processing phases.
Step 2 Standardize to mono audio clips. Standardizing multitrack or variable audio files into single-channel mono audio uniformizes inputs. This streamlines data dimensionality, prevents balance bias, and guarantees consistent array structures across a machine learning pipeline.
Step 3 Use an Inverse Fourier transform on frequency changes over time. Utilizing an Inverse Fourier transform converts processed spectrum components back into the time domain, mapping frequency-specific variations accurately across time for downstream analysis.

Question #2 (Topic: demo questions)

DRAG DROP You need to define a process for penalty event detection. Which three actions should you perform in sequence? To answer, move the appropriate actions from the list of actionsto the answer area and arrange them in the correct order.

A.

Actions (Available Options)
Build the global model using Microsoft Cognitive Toolkit (CNTK).
Import the global model and build a local model using Microsoft Cognitive Toolkit (CNTK).
Export the global model using Neural Network Exchange Format (NNEF).
Import the global model and build the local model using PyTorch.
Build the global model using PyTorch.
Build the global model using TensorFlow.
Import the global model and build the local model using TensorFlow.
Export the global model using the Open Neural Network Exchange (ONNX) format.
Answer Area (Ordered Steps)
Step Sequence Action to Perform
Step 1 Build the global model using PyTorch.
Step 2 Export the global model using Neural Network Exchange Format (NNEF).
Step 3 Import the global model and build the local model using TensorFlow.

Correct Answer: A

Explanation:

Step Action Technical Significance
Step 1 Build the global model using PyTorch. The machine learning pipeline initiates by developing and training a foundational core representation (the global model) using a dynamic framework like PyTorch.
Step 2 Export the global model using Neural Network Exchange Format (NNEF). To port or hand off the model across heterogeneous environments, software layers, or edge frameworks, the global model structure is serialized into an open, cross-platform architecture such as NNEF.
Step 3 Import the global model and build the local model using TensorFlow. The standardized architecture is then pulled or imported into another production environment (TensorFlow) to adjust parameters locally or execute fine-tuning for deployment scenarios.

Question #3 (Topic: demo questions)

You need to select an environment that will meet the business and data requirements. Which environmentshould you use?

A.

Azure HDInsight with Spark MLlib

B.

Azure Cognitive Services

C.

Azure Machine Learning Studio

D.

Microsoft Machine Learning Server

Correct Answer: D

Explanation:

Microsoft Machine Learning Server is the best choice when an organisation needs to build and deploy machine learning solutions while meeting strict business and data requirements. It can run on-premises and process data where it resides, which is important for organisations that cannot move sensitive or large volumes of data to the cloud due to compliance, security, or regulatory requirements. It supports scalable analytics using R and Python and can integrate with existing enterprise systems. In contrast, Azure Cognitive Services provides prebuilt AI APIs, Azure Machine Learning Studio is primarily a cloud-based machine learning development environment, and Azure HDInsight with Spark MLlib focuses on big data processing rather than meeting enterprise data residency and on-premises requirements. Therefore, Microsoft Machine Learning Server is the most appropriate solution.

Question #4 (Topic: demo questions)

You need to resolve the local machine learning pipeline performance issue. What should you do?

A.

Increase Graphic Processing Units (GPUs).

B.

Increase the learning rate.

C.

Increase the training iterations,

D.

Increase Central Processing Units (CPUs).

Correct Answer: A

Explanation:

If the local machine learning pipeline is experiencing performance issues, increasing the number of GPUs is the most effective solution. GPUs are specifically designed to perform the large-scale parallel computations required for machine learning and deep learning workloads. They can process thousands of operations simultaneously, significantly reducing model training and inference times. In contrast, increasing the learning rate or training iterations affects the model's learning process rather than performance, and adding CPUs generally provides less improvement for machine learning workloads than GPUs. Therefore, increasing GPUs is the best way to resolve the pipeline performance issue.

Page: 1 / 1