Datasets in PyTorch QUIZ (MCQ QUESTIONS AND ANSWERS)

download() method

fetch() method

torchvision.datasets.load()

torch.datasets.get()

Replace missing values with the mean of the dataset

Ignore samples with missing values

Impute missing values using a machine learning model

Assign a default value to missing data

Use the torch.set_seed() function

Set the learning rate to a fixed value

Increase the number of workers in the DataLoader

Use a larger batch size during training

A single forward pass of the neural network

The entire training dataset passed through the model once

The number of layers in the neural network

The learning rate at a specific iteration

Concatenating different datasets into a single dataset

Concatenating layers of a neural network

Concatenating data along the time axis

Concatenating batches during training

torch.show()

plt.imshow() from the matplotlib library

print()

torchvision.display()

Dropping the last batch if it is incomplete

Dropping the last epoch during training

Ignoring the last sample in the dataset

Removing the last layer of the neural network

To randomly shuffle the dataset

To split the dataset into training and validation sets

To apply random transformations to the data

To create a DataLoader with random sampling

Use a larger learning rate for minority classes

Oversample the minority classes

Undersample the majority classes

Ignore the imbalances and let the model adapt

To define neural network architectures

To apply data augmentation and preprocessing transformations

To calculate loss functions

To create custom DataLoader objects

For every machine learning project

When working with small, specific datasets

When using image classification tasks only

When computational resources are limited

Faster convergence

Improved memory efficiency

Increased GPU memory usage

More accurate model predictions

To specify the number of GPU devices

To set the learning rate

To control data shuffling

To determine the number of parallel data loading processes

torch.nn

torch.optim

torch.utils.data

torchvision.models

Efficient GPU utilization

Faster convergence

Automated hyperparameter tuning

Improved memory management

Showcase PyTorch's capabilities

Provide convenient access to common datasets for research

Replace custom datasets entirely

Facilitate communication between different frameworks

Image classification with 100 classes

Object detection in 100 images

Natural language processing tasks

Regression analysis on 100 features

MNIST

CIFAR-100

ImageNet

TensorFlow MNIST

Min-Max scaling

One-Hot encoding

Z-score normalization

Principal Component Analysis (PCA)

To control the learning rate

To specify the number of epochs

To apply data augmentation and preprocessing

To determine the loss function

pandas library

numpy library

Image processing libraries (e.g., PIL)

All of the above

Overfitting

Model convergence

Data leakage

Stuck in local minima

Data augmentation

Model training

GPU acceleration

Data shuffling and batching

To define the length of the dataset

To specify the number of features in each sample

To control data augmentation techniques

To set the learning rate during training

__init__

__len__

__getitem__

__load__

torch.Dataset

torch.BaseDataset

torch.CustomDataset

torch.DataLoader

Improved efficiency

Greater control over data preprocessing

Faster training times

Automatic data augmentation

Use the torch.info() function

Check the dataset's documentation

Use the len() function on the dataset

Print the dataset object

They are more efficient than custom datasets

They are pre-trained on large datasets

They provide a standardized way to access common datasets

They are exclusive to PyTorch and not available in other frameworks

torch.data

torch.datasets

torch.utils

torch.loader