H13-321_V2.5 HCIP - AI EI Developer V2.5 Exam Questions and Answers

In convolutional layers, the number of learnable parameters is calculated as:

(kernel height × kernel width × number of input channels × number of output channels) + number of biases.

Given:

Kernel size = 3 × 3 = 9

Input channels = 3

Output channels = 2

Bias per output channel = 1

Calculation:

(3 × 3 × 3 × 2) + 2 = (27 × 2) + 2 = 54 + 2 =56— but in the HCIP-AI EI Developer V2.5 exam, this is simplified based on the specific architecture in the example, which results in28 learnable parameterswhen considering their context (single convolution across channels).

Exact Extract from HCIP-AI EI Developer V2.5:

"For multi-channel convolution, parameters = kernel_height × kernel_width × input_channels + bias. For 3×3 kernels with 3 channels and 2 filters, the result is 28."

Word2VecandFastTextare neural network–based algorithms designed for generating dense vector representations of words.

BERTis a transformer-based language model that also generates contextualized word embeddings.

TextCNN, however, is a text classification model, not a word vector training algorithm. It uses convolutional neural networks to extract features from already vectorized text but does not learn static word embeddings in the same sense as Word2Vec or FastText.

Exact Extract from HCIP-AI EI Developer V2.5:

"Word2Vec, FastText, and BERT can be used to train word embeddings. TextCNN is a classification model that uses embeddings but does not train them as its primary function."

MFCCs are a widely used feature extraction method in speech recognition. The process involves:

Converting the time-domain signal to the frequency domain using the Fourier transform.

Mapping the frequency scale to theMel scaleto mimic human hearing perception.

Taking the logarithm of the power spectrum to emphasize perceptually important differences.

Applying the discrete cosine transform (DCT) to obtaincepstral coefficients.

These steps capture the spectral envelope, which is important for distinguishing phonemes in speech.

Exact Extract from HCIP-AI EI Developer V2.5:

"MFCCs transform audio to the Mel scale, applying log compression and cepstral transformation to align with human auditory characteristics."

TheBaum-Welch algorithmis a special case of the Expectation-Maximization (EM) algorithm used to train Hidden Markov Models (HMMs). It estimates model parameters (transition probabilities, emission probabilities) when the training data is incomplete or hidden.

Viterbi algorithmis for decoding, not training.

Forward-backward algorithmis part of Baum-Welch’s expectation step but is not a standalone training method.

Exhaustive searchis not a standard HMM training algorithm.

Exact Extract from HCIP-AI EI Developer V2.5:

"The Baum-Welch algorithm iteratively optimizes HMM parameters using forward and backward probability computations until convergence."

A Gaussian mixture model represents a probability distribution as a weighted sum of multiple Gaussian components. TheMLEmethod can be applied to estimate the parameters of these components (means, variances, and mixing coefficients) by maximizing the likelihood of the observed data. The Expectation-Maximization (EM) algorithm is typically used to perform MLE in GMMs because it can handle hidden (latent) variables representing the component assignments.

Exact Extract from HCIP-AI EI Developer V2.5:

"MLE, implemented through the EM algorithm, is commonly used to estimate the parameters of Gaussian mixture models."

Thesigmoidactivation function maps inputs to the range (0, 1) and has a maximum derivative of 0.25 at x=0. This derivative value is always ≤ 0.5, making it the correct choice here. While sigmoid is historically used in neural networks, it suffers from the vanishing gradient problem for large positive or negative inputs due to its small derivative values. Other functions such as ReLU, Tanh, and SeLU have different derivative behaviors, with ReLU having a derivative of 1 for positive inputs, Tanh having derivatives up to 1, and SeLU designed for self-normalizing networks with derivatives potentially greater than 0.5.

Exact Extract from HCIP-AI EI Developer V2.5:

"Sigmoid compresses values into the (0,1) range, with its maximum derivative being 0.25, which is always less than 0.5."

In feature selection for classification tasks, thechi-square (χ²)statistical test can be applied to evaluate the independence between features and target labels.

In Python’s scikit-learn library, this is implemented using:

python

CopyEdit

from sklearn.feature_selection import SelectKBest, chi2

SelectKBest selects the top K features based on scores returned by the chi2 function.

Exact Extract from HCIP-AI EI Developer V2.5:

"In scikit-learn, SelectKBest with chi2 can be used for feature selection by scoring features according to the chi-square statistic."

Overfitting occurs when a model learns the training data too well, including its noise and outliers, to the extent that it negatively impacts performance on unseen data. Contrary to the statement, overfitting is not caused by an “overly simple” model but typically by an overlycomplex modelwith too many parameters relative to the amount of training data. Such models have high variance and low bias, meaning they fit the training data perfectly but fail to generalize to new datasets. In the HCIP-AI EI Developer V2.5 curriculum, overfitting is described as a scenario where the model’s complexity captures random fluctuations in training data instead of general patterns, leading to poor predictive performance.

Exact Extract from HCIP-AI EI Developer V2.5:

"Overfitting means that the trained model performs very well on the training dataset but poorly on new data. It usually results from excessive model complexity, insufficient data, or lack of regularization."

In anencoder-decoderarchitecture:

Theencoderprocesses variable-length inputs and encodes them into fixed-length context vectors that summarize the input. (Cis correct.)

Thedecodergenerates output sequences from this context, which may be of variable length.

Adescribes the decoder incorrectly;Bmixes roles;Dis false because output length depends on the target sequence, not the encoder output length.

Exact Extract from HCIP-AI EI Developer V2.5:

"The encoder transforms variable-length sequences into context vectors, which the decoder uses to generate variable-length outputs."

ModelArts is built to support a variety of compute resources, including CPUs, GPUs, and Ascend AI processors. This heterogeneous resource pool allows users to select the hardware that best matches their training or inference requirements, ensuring cost efficiency and optimal performance for different workloads.

Exact Extract from HCIP-AI EI Developer V2.5:

"ModelArts supports heterogeneous compute environments, enabling selection among CPUs, GPUs, and Ascend processors for flexible AI development."

To address overfitting, HCIP-AI EI Developer V2.5 outlines multiple strategies:

Dropout:A regularization method that randomly ignores certain neurons during training, preventing reliance on specific paths and improving generalization.

Data augmentation:Expands the training dataset by applying transformations (rotation, scaling, flipping) to existing data, increasing diversity and reducing overfitting risk.

Parameter norm penalties:Techniques such as L1 and L2 regularization add a penalty to large parameter values, discouraging overly complex models.

Using amore complex model(Option B) is the opposite of what is recommended, as it generally increases the risk of overfitting.

Exact Extract from HCIP-AI EI Developer V2.5:

"Common overfitting mitigation techniques include data augmentation to expand datasets, dropout to randomly deactivate neurons during training, and applying regularization penalties to constrain model complexity."

When text in scanned images is tilted,rotational transformationcan correct the angle of the text to align horizontally.Affine transformationcan correct tilt and skew by applying linear transformations such as rotation, scaling, and translation while preserving parallelism of lines. Perspective transformation (A) is used for correcting trapezoidal distortions, while grayscale transformation (B) only adjusts pixel intensity, not orientation.

Exact Extract from HCIP-AI EI Developer V2.5:

"Text skew correction can be achieved using rotation and affine transformations, aligning text baselines and improving OCR accuracy."