B.TechSemester 72024-25Deep LearningKCS078

Deep Learning (KCS078) - AKTU Question Paper 2024-25

B.Tech · Semester 7 · Free PDF Download

This is the official AKTU Deep Learning Previous Year Question Paper for B.Tech Semester 7, academic session 2024-25. Published by Dr. A.P.J. Abdul Kalam Technical University (AKTU/UPTU), Lucknow. Free PDF download — no login required.

Course:B.Tech

Semester:Semester 7

Session:2024-25

University:AKTU / UPTU

Rate this paper

Questions Asked in 2024-25

Deep Learning (KCS078) — complete question paper · 100 marks · 3 Hours

Section AAttempt all questions in brief. 2 x 10 = 20

a
Discuss the differences between a perceptron and a support vec tor machine in terms of decision boundary formation
b
Discuss how the choice of a loss function affects model perfor mance for
a
classification problem
c
Trace the historical development of deep learning that enabled d e e p networks to outperform traditional machine learning models
d
How does weight sharing in C NNs improve efficiency and performance?
e
Explain the impact of weight initialization on the performance of deep networks
f
Compare performance of autoenco ders with traditional dimension ality reduction techniques
g
Provide examples of optimization algorithms used for deep netw orks. 4 K 2
h
How do techniques address the challenges of stochastic optimiz ation in deep learning? i. Describe the architecture and training of WaveNet for audio generation. 5 K 3, j. Discuss the role of Word2Vec i n natural language processing. 5 K 3

Section BAttempt any three of the following: 10 x 3 = 30

a
Derive the gradient of the logistic regression loss function with respect to its parameters. How does regularization impact its optimizat ion process?
b
Define VC dimension and disc uss its implications for understanding the generalization capabilities of deep networks
c
Explain the role of manifold l earning in dimensionality reduction. How does it compare to linear methods?
d
Explain the concept of generalization in neural networks. Ho w do techniques like dropout and w eight regularization improve generalization performance?
e
Discuss the pipeline of scene understanding using deep learn ing models. How does this pipeline handle challenges like occlusion and var ying lighting conditions?

Section CAttempt any one part of the following: 10 x 1 = 10

a
Prove that neural networks ca n act as universal function approximators. Discuss any constraints or assumptions involved in this theorem
b
Discuss the relationship betw een the choice of activation fu nction in a shallow neural network and its ability to model non-linear relationships
a
Analyze the trade-offs between increasing network depth and the risk of vanishing/exploding gradients. Wha t role does skip connections in ResNet play in addressing these issues?
b
How does semi-supervised learning differ from supervised and unsupervised learning? Discuss its application in training deep networks with limited labeled data
a
Explain the concept of metric learning in the context of dim ensionality reduction. How does it differ from traditional techniques like PCA
b
Analyze the architectural differences between AlexNet, VGG, Inception, and ResNet. How do these architectures address the vanishing gradient problem?
a
Discuss the implications of non-convex optimization in desig ning deep learning models. How do saddle points influence the training process?
b
Provide a detailed overview o f computational neuroscience pr inciples applied in designing artificial neural networks
a
Provide a detailed overview of deep learning-based technique s for audio detection. How does WaveNet outperform traditional methods?
b
Describe the process of joint detection and captioning in im age captioning systems. How does attention enhance this process?

Question text is extracted from the official AKTU question paper PDF above. Hindi translations are omitted — every question is printed in English in the original paper. Last verified: 2026-08-23.

Deep Learning — Other Year Papers

AKTU Deep Learning PYQs from other sessions

Deep Learning2022-23 Deep Learning2024-25 Deep Learning2024-25 Deep Learning2025-26

More B.Tech Semester 7 (2024-25) Papers

Other subjects from same semester and session

Electric Drives2024-25 Additive Manufacturing2024-25 Machine Learning2024-25 Maintenance Engineering Management2024-25 Disaster Preparedness And Management2024-25 Speech Processing2024-25 Power Plant Engineering2024-25 Vlsi Design2024-25

View all B.Tech Semester 7 2024-25 papers →

Syllabus & More PYQs

Paper solve karne se pehle unit-wise syllabus dekh lo

AKTU 4th Year PYQs — All Branches