NCA-GENL Core Machine Learning and AI Knowledge Practice Question
Exhibit
config.json:
{
"optimizer": "Adam",
"learning_rate": 0.01,
"batch_size": 32,
"precision": "FP32"
}
training_log:
Epoch 1: loss = 2.45
Epoch 2: loss = 2.38
Epoch 3: loss = 2.42
Epoch 4: loss = 2.39Refer to the exhibit. The training loss is oscillating and failing to converge. What is the most likely immediate adjustment needed?
⚠ Common exam trap
Many candidates assume that an oscillating loss requires increasing the training epochs or batch size, overlooking the primary symptom of an excessively high learning rate.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Reduce the learning rate
The oscillating loss and lack of convergence suggest the learning rate is too high. A learning rate of 0.01 is relatively aggressive for many deep learning tasks, causing the model to jump over the optimal minima. Reducing the learning rate is the standard first step to stabilize the training process, allowing the optimizer to settle into the local minimum more consistently and achieve better overall convergence results.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Reduce the learning rate
Why this is correct
An oscillating loss is a classic symptom of an overly aggressive learning rate. By reducing the learning rate, the optimizer makes smaller updates to the weights, preventing the model from 'bouncing' around the loss landscape. This allows for finer adjustments and more stable convergence toward a lower loss value.
- ✗
Increase the batch size to 1024
Why it's wrong here
While increasing batch size can stabilize gradients, it does not address the fundamental issue of the learning rate being too high. Furthermore, changing batch size significantly requires tuning the learning rate as well. If the learning rate remains at 0.01, a larger batch size may still lead to oscillations.
- ✗
Switch to a larger model architecture
Why it's wrong here
Switching to a larger model will likely increase the instability if the current learning rate is already too high. Larger models have more parameters that can react violently to large gradient updates. The issue is clearly optimization-related, and increasing complexity will only exacerbate the problems with finding a stable minimum.
- ✗
Switch to FP16 mixed-precision
Why it's wrong here
Switching to FP16 might offer speed benefits and memory savings, but it does not address the convergence instability caused by the learning rate. In fact, if the training is unstable, FP16 might introduce further precision-related issues if not paired with appropriate loss scaling. The primary problem remains the optimization hyperparameter, not data precision.
About these practice questions
Courseiva writes every NCA-GENL question from scratch — 367 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official NVIDIA exam blueprint
This NCA-GENL practice question is part of Courseiva's free NVIDIA certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the NCA-GENL exam.