GNRS 558 ACTUAL TEST PAPER 2026 COMPLETE
QUESTIONS AND CORRECT ANSWERS GRADED
A+
▶ Without considering any regularization terms, the support vectors of a
support vector machine (SVM) are composed of ( ).
A. Points on the separating hyperplane
B. The points farthest from the separating hyperplane
C. The points closest to the separating hyperplane
D. Points of a certain type - Answer: C
▶ Within the broad landscape of large language models built on the
Transformer architecture, which roadmap does GPT belong to?
A. Encoder-Only
B. Decoder-Only
C. Encoder-Decoder
D. Decoder-Encoder - Answer: B
▶ Which of the following is NOT a use case for natural language
processing?
A. Public opinion analysis
B. Machine translation
C. Text classification
D. Image recognition - Answer: D
▶ Which of the following statements about different optimizers is false?
A. RMSprop solves the problem of the Adagrad optimizer ending too early.
B. Compared with RMSprop, Adagrad is more sensitive to gradient
changes.
C. Both Adagrad and RMSprop can set adaptive learning rate for
parameters.
D. Adam requires bias correction for initial iterations. - Answer: B
▶ Which of the following is NOT an Al deep learning framework?
A. TensorFlow
B. Scikit-learn
C. Pytorch
QUESTIONS AND CORRECT ANSWERS GRADED
A+
▶ Without considering any regularization terms, the support vectors of a
support vector machine (SVM) are composed of ( ).
A. Points on the separating hyperplane
B. The points farthest from the separating hyperplane
C. The points closest to the separating hyperplane
D. Points of a certain type - Answer: C
▶ Within the broad landscape of large language models built on the
Transformer architecture, which roadmap does GPT belong to?
A. Encoder-Only
B. Decoder-Only
C. Encoder-Decoder
D. Decoder-Encoder - Answer: B
▶ Which of the following is NOT a use case for natural language
processing?
A. Public opinion analysis
B. Machine translation
C. Text classification
D. Image recognition - Answer: D
▶ Which of the following statements about different optimizers is false?
A. RMSprop solves the problem of the Adagrad optimizer ending too early.
B. Compared with RMSprop, Adagrad is more sensitive to gradient
changes.
C. Both Adagrad and RMSprop can set adaptive learning rate for
parameters.
D. Adam requires bias correction for initial iterations. - Answer: B
▶ Which of the following is NOT an Al deep learning framework?
A. TensorFlow
B. Scikit-learn
C. Pytorch
