Assignment 01
Deep Learning Essentials
Instructions:
Answer the following questions based on the concepts covered in the lectures. Use
external resources to research (i.e. Google) to explore and expand your
understanding. Each question should be answered in 150–200 words.
Question 1: Exploring Neural Network Architectures
In the lecture on Neural Networks, we learned about fully connected neural
networks and their application in house price prediction. However, there are more
advanced neural network architectures beyond fully connected networks. Research
and describe two advanced neural network models (e.g., Convolutional Neural
Networks, Recurrent Neural Networks, or Transformers). Explain their unique
features, how they differ from fully connected networks, and provide one
real-world application for each.
Question 2: Beyond Sigmoid: Activation Functions in Neural
Networks
The lecture on Logistic Regression introduced the sigmoid function as a key
component for binary classification. However, there are many other activation
functions used in neural networks. Research and describe two additional
activation functions (e.g., ReLU, Tanh, or Leaky ReLU). Explain how they work,
their advantages over the sigmoid function, and in what scenarios they are
commonly used.
Question 3: Exploring Loss Functions
In the lecture on Loss Functions, we discussed the negative log loss (binary
cross-entropy) as a common loss function for binary classification. However,
different tasks require different loss functions. Research and describe two other
loss functions (e.g., Mean Squared Error, Hinge Loss, or Cross-Entropy Loss for
multi-class classification). Explain their mathematical formulation, the type of
problems they are used for, and why they are suitable for those tasks.
Bonus Activity: Interactive Practice
Visit the TensorFlow Playground (https://playground.tensorflow.org/)and experiment
with building a simple neural network. Try changing the activation functions, adding
layers, and observing how the model learns. Write a short reflection (50–100 words)
on how changing these parameters affects the model’s performance.
TensorFlow Playground
Submission Guidelines:
● Submit your answers in a PDF document.
● Include references for any external resources you used.
● The bonus activity reflection can be included at the end of your document.
Grading Criteria
Each question is worth 10 marks, and the assignment is graded out of a total of 30
marks. The grading will be based on the following criteria:
1. Accuracy and Depth of Explanation (4 marks):
○ The answer should accurately explain the concepts and provide
sufficient depth.
○ Use of relevant examples or applications to support the explanation.
2. Clarity and Organization (3 marks):
○ The answer should be well-structured, clear, and easy to follow.
○ Proper use of paragraphs, headings, and bullet points (if applicable).
3. Research and Exploration (2 marks):
○ Evidence of research beyond the lecture content.
○ Use of credible external resources to enhance the explanation.
4. Originality and Reflection (1 mark):
○ The answer should demonstrate original thought and personal
reflection.
○ Avoid copying directly from sources; paraphrase and synthesize
information.
Total Marks: 30
● Question 1: 10 marks
● Question 2: 10 marks
● Question 3: 10 marks
The bonus activity is optional and does not contribute to the grading but is highly
recommended for a better understanding of the concepts.
Good luck!
