NVIDIA Generative AI Multimodal 認定 NCA-GENM 試験問題:

1. You are training a Variational Autoencoder (VAE) and notice that the generated samples are blurry and lack detail. Which of the following adjustments could help improve the quality and sharpness of the generated images2 Select all that apply.

A) Decrease the batch size to reduce computational complexity
B) Use a more powerful decoder architecture, such as one with deconvolutional layers.
C) Increase the capacity of the encoder and decoder networks by adding more layers or units.
D) Increase the dimensionality of the latent space
E) Decrease the weight of the Kullback-Leibler (KL) divergence term in the loss function-

2. You observe that the generated images often lack fine-grained details and tend to be blurry. Which of the following techniques could MOST effectively improve the visual quality of the generated images?

A) Using a variational autoencoder (VAE) instead of a GAN.unlikely to significantly improve diagnosis accuracy.
B) Increasing the batch size during training.
C) Decreasing the learning rate during training.
D) Implementing a discriminator network and using adversarial training (GAN).
E) Using a larger dataset of text-image pairs.

3. You are tasked with creating a multimodal AI application that analyzes social media posts containing text, images, and user profile information to predict the likelihood of a post going viral. Which feature engineering techniques are most effective for representing and integrating these different modalities?

A) Using a combination of TF-IDF for text, pixel values for images, and numerical features for user profile information. Then apply PCA for dimensionality reduction.
B) Using bag-of-words for text, histogram of oriented gradients (HOG) for images, and simple numerical features (e.g., number of followers) for user profiles.
C) Using TF-IDF for text, pixel values for images, and one-hot encoding for user profile information.
D) Using word embeddings (e.g., Word2Vec, GloVe) for text, pre-trained CNN features (e.g., from ResNet, Inception) for images, and embedding user profiles using a graph embedding technique.
E) Using character-level n-grams for text, edge detection for images, and boole an features for user profile information.

4. You're training a Generative Adversarial Network (GAN) to generate images from text descriptions. After a few epochs, you notice the generator is producing nearly identical images regardless of the text input (mode collapse). Which of the following strategies could help mitigate this issue?

A) Apply weight decay regularization to the generator
B) Implement mini-batch discrimination in the discriminator.
C) Use a larger batch size for the generator.
D) Increase the capacity (number of layers/neurons) of the discriminator.
E) Decrease the learning rate of the discriminator.

5. Given the following Python code snippet using Pandas, which is intended to filter rows where the 'price' column is greater than 100 and the 'quantity' column is less than 5, identify the correct approach to achieve this:

A)

B)

C)

D)

E)

質問と回答：