Fundamentals of Machine Learning II (Not for Certification)

Question 1

What is deep learning?

Answer 1

Deep learning is an advanced form of machine learning that uses artificial neural networks with multiple layers to model complex patterns in data, mimicking the way the human brain processes information.

Question 2

What is an artificial neural network (ANN)?

Answer 2

An ANN is a computational model inspired by biological neural networks, composed of interconnected layers of nodes (neurons) that process input data through mathematical functions to produce an output.

Question 3

Why is it called “deep” learning?

Answer 3

The term “deep” refers to the multiple layers of neurons in a deep neural network, creating a deeply nested function.

Question 4

What types of problems can deep learning be used for?

Answer 4

Deep learning can be used for regression, classification, natural language processing, and computer vision.

Question 5

What is a neuron in a neural network?

Answer 5

A neuron is a single computational unit in a neural network that applies a mathematical function to its inputs and passes the result to the next layer based on an activation function.

Question 6

hat is the role of weights (w) in a neural network?

Answer 6

Weights (w) determine how much influence each input (x) has on the output. They are adjusted during training to reduce prediction errors.

Question 7

What is an activation function?

Answer 7

An activation function determines whether the output of a neuron should be passed to the next layer. Examples include ReLU, sigmoid, and softmax.

Question 8

What is a classification problem in deep learning?

Answer 8

A classification problem is where a model predicts the probability of an input belonging to a certain class, such as predicting penguin species based on features.

Question 9

How does a neural network predict class probabilities?

Answer 9

The output layer produces a vector of probabilities for each class using functions like softmax. The class with the highest probability is selected as the prediction.

Question 10

What is the loss function in deep learning?

Answer 10

A loss function measures the error between predicted values (ŷ) and true values (y). The goal is to minimize this loss during training.

Question 11

What is backpropagation?

Answer 11

Backpropagation is the process of adjusting weights in a neural network by propagating the loss backward through the layers to reduce error.

Question 12

What optimization technique is commonly used to adjust weights in deep learning?

Answer 12

Gradient descent is commonly used, adjusting weights up or down to minimize the loss function by finding the direction of maximum decrease.

Question 13

How are training data processed in neural network training?

Answer 13

Training data are batched into matrices and processed using linear algebraic calculations, which is why GPUs are used for deep learning tasks.

Question 14

What is an epoch in the context of neural network training?

Answer 14

An epoch refers to one complete pass through the entire training dataset during the learning process, with weights being adjusted after each pass.

Question 15

What is the softmax function?

Answer 15

Softmax is a type of activation function used in the output layer of a neural network for classification tasks. It converts output scores into probabilities for each class.

Question 16

What is the purpose of Azure Machine Learning?

Answer 16

It helps manage the entire lifecycle of machine learning projects, including data exploration, model training, deployment, and monitoring for responsible AI practices.

Question 16

What is Microsoft Azure Machine Learning?

Answer 16

Microsoft Azure Machine Learning is a cloud service that provides tools for training, deploying, and managing machine learning models, designed for data scientists, software engineers, and DevOps professionals.

Question 17

What are the key stages in the Azure Machine Learning project lifecycle?

Answer 17

1. Exploring and preparing data.
2. Training and evaluating models.
3. Registering and managing models.
4. Deploying models for use.
5. Reviewing responsible AI principles.

Question 17

What are some features provided by Azure Machine Learning for machine learning workloads?

Answer 17

Key features include centralized dataset management, on-demand compute resources, automated machine learning (AutoML), visual tools for pipeline orchestration, and integration with frameworks like MLflow.

Question 18

How does Azure Machine Learning support responsible AI?

Answer 18

It provides built-in tools for visualizing and evaluating metrics like model explainability, fairness, and bias assessment to ensure responsible AI practices are followed.

Question 18

What is Automated Machine Learning (AutoML)?

Answer 18

AutoML is a feature in Azure Machine Learning that automates the process of running multiple training jobs with different algorithms and parameters to find the best model for a dataset.

Question 19

What are visual tools in Azure Machine Learning used for?

Answer 19

They are used to define orchestrated pipelines, which automate processes like model training, inferencing, and batch processing tasks.

Question 20

What is an Azure Machine Learning workspace?

Answer 20

An Azure Machine Learning workspace is the central resource needed for organizing and managing all machine learning assets, such as datasets, models, and compute resources, within Azure.

Question 21

How do you provision an Azure Machine Learning workspace?

Answer 21

It can be provisioned through the Azure portal, and it automatically creates supporting resources like storage accounts, container registries, and virtual machines as needed.

Question 22

What is Azure Machine Learning Studio?

Answer 22

Azure Machine Learning Studio is a browser-based portal for managing machine learning resources, running jobs, and accessing tools like notebooks and visual pipelines.

Question 22

What are some tasks you can perform in Azure Machine Learning Studio?

Answer 22

Tasks include importing data, creating compute resources, running code in notebooks, using AutoML, creating pipelines, and deploying models for inferencing.

Question 23

How does Azure Machine Learning handle model deployment?

Answer 23

It supports both on-request (real-time) and batch inferencing, allowing trained models to be deployed for use by applications and services.

Question 24

What is the role of on-demand compute resources in Azure Machine Learning?

Answer 24

On-demand compute resources provide scalable virtual machines or clusters that can be used to run machine learning jobs, such as model training and evaluation.

Question 25

How does Azure Machine Learning integrate with other machine learning frameworks?

Answer 25

It integrates with frameworks like MLflow, which allows for easier management of model training, evaluation, and deployment at scale.

Question 26

What are pipelines in Azure Machine Learning?

Answer 26

Pipelines are orchestrated sequences of tasks that automate the process of running machine learning jobs, such as data preprocessing, model training, and deployment.

Question 27

What is the model catalog in Azure Machine Learning?

Answer 27

The model catalog is a comprehensive collection of pre-trained and user-created machine learning models, which can be imported, managed, and used for deployment.

Question 28

What responsible AI principles are integrated into Azure Machine Learning?

Answer 28

Responsible AI principles include fairness, transparency, accountability, and privacy, which are supported through features like model explainability and fairness assessments.

Question 29

How does Azure Machine Learning assist in model evaluation?

Answer 29

It provides tools for evaluating models using metrics, tracking performance, and assessing key areas such as accuracy, precision, and recall, along with responsible AI metrics.

Question 30

What are storage accounts in Azure Machine Learning used for?

Answer 30

Storage accounts are used to store datasets, experiment results, and models required for machine learning projects.

Question 31

Why are GPUs recommended for Azure Machine Learning tasks?

Answer 31

GPUs (Graphics Processing Units) are optimized for vector and matrix calculations, making them ideal for speeding up large-scale machine learning tasks like model training and inferencing.