---
source_file: "Deep Learning - UTS - 2022 - IUP.pdf"
source_type: "born-digital PDF (1 page) with embedded table image"
transcription_method: "embedded text layer (pypdf) + visual reading of the embedded table image"
course: "Deep Learning IUP Computer Science"
exam: "MidTerm Exam (UTS) — Midterm"
term: "Even Semester 2021/2022"
program: "Program Studi S1 Ilmu Komputer (IUP — International Undergraduate Program)"
faculty: "Fakultas Matematika dan Ilmu Pengetahuan Alam (FMIPA)"
university: "Universitas Gadjah Mada (UGM)"
date: "Monday, 4 April 2022"
duration: "90 menit (90 minutes)"
lecturer: "Afiahayati, S.Kom., M.Cs., Ph.D."
exam_type: "Open book and Online"
pages: 1
topics: [perceptron, multi-layer-perceptron, backpropagation, vanishing-gradient, overfitting, CNN]
---

# UTS Deep Learning IUP — 2021/2022

> **Header (as printed):**
> PROGRAM STUDI S1 ILMU KOMPUTER
> FAKULTAS MATEMATIKA DAN ILMU PENGETAHUAN ALAM
> UNIVERSITAS GADJAH MADA
>
> MidTerm Exam
> Even Semester 2021/2022

| Field | Value |
|---|---|
| Course | Deep Learning IUP Computer Science |
| Day, date | Monday, 4 April 2022 |
| Time | 90 menit |
| Lecturer | Afiahayati, S.Kom., M.Cs., Ph.D. |
| Exam | Open book and Online |

**Instructions:**
Please write the answers in a paper/papers, then take picture of the papers and upload to eLok/simaster.

---

## 1. CO 1 — Able to differentiate Deep Learning to traditional Neural Network (7.5 %)

a) Please explain the difference of learning algorithm in Multi-layer Perceptron and Single Layer Perceptron!

b) Which one is better Multi Layer Perceptron or Single Layer Perceptron? Please explain!

c) Please explain the differences between deep learning and traditional neural network!

---

## 2. CO 2 — Able to understand multi-layer perceptron and backpropagation (7.5 %)

Please pay attention to the below table. There are two attributes (X₁ and X₂) to classify class/label (y). If y=0 then class=0 else if y=1 then class=1

| no | X₁ | X₂ | y |
|---|---|---|---|
| 1 | 5 | 6 | 1 |
| 2 | 2 | 1 | 0 |
| 3 | 7 | 8 | 1 |
| 4 | 2 | 3 | ??? |

Please give an architecture illustration of multi-layer perceptron with 1 hidden layer consisting of 3 neurons to develop a classification model for the above table!

---

## 3. CO 3 — Able to understand components in DNN architecture (10%)

a) Backropagation algorithm will cause vanishing gradient in deep neural network. Please explain what is the vanishing gradient!

b) There are 3 strategies from deep learning to handle the vanishing gradient, please explain the strategies!

c) Please explain how to detect that our classification model is overfit!

d) Please explain two strategies to tackle overfit!

---

## 4. CO 5 — Able to understand DNN architectures including CNN (modern/traditional) and RNN/LSTM/GRU (5%)

a) Please explain the architecture of CNN!

b) Please give case examples which can be solved by CNN!

---

Tanda tangan Reviewer:
Telah direview pada tanggal: ……………..
Nama Reviewer: …………………………..