You are on page 1of 2

National University of Sciences & Technology

School of Electrical Engineering and Computer Science


Department of Computing

Task: Find a good paper that could be improved

Maximum Marks: 30 Instructor: Dr. Khuram Shahzad

Submission Date: Thursday 3rd march Type: Written / Individual (later form groups)

Name: salman arif Reg. 333781 Degree: BSCS Section:10A

Instructions:

1. Go to https://paperswithcode.com/ https://github.com/zziz/pwc, https://researchcode.com/,


https://paperswithcode.com/sota]and search for a good paper related to your area of interest using key words
2. A good paper can be found based on; 1) have at least one code implementations preferably in python, 2) not
more than three years old, 3) limitations and improvement are clearly mentioned, 4) good venue (published
journal with impact factor, IEEE, Springer, Elsevier, MDPI etc.), 5) have good citations that could be confirmed
from www.scholar.google.com by search paper with title.
3. Limitations and improvement can be found from abstract, introduction, assumptions, and concussions and
future work sections; you do not need to read entire paper for this assignment.
4. If task (3) is not clearly given in a paper, do step 1-3 again unless you find a good paper that could clearly be
improved.
5. Next Task 2 is to run the code, if you could do not in assignment 1 it would save time (Assignment 2).
6. Write ½-1 page in listed form; limitations and improvements (table below)
7. In filling table below, try to link limitation to a corresponding improvement with method name, e.g., one
improvement could be; Use sequential Bayesian methods [13] (method to solve a problem) for better order
book reconstruction and estimation of order queues (limitation) for a paper.
8. After running code, attach screen shots.
9. Journal paper details

Paper Title Journal Impact Citations Publication


Name factor year

Neural networks for stock price prediction Journal of 1.476 Y.S. Abu- May 30, 2018
Difference Mostafa and
Equations and A.F. Atiya,
Applications Introduction
to _nancial
forecasting,
Applied Intel-
ligence
6.3(1996), pp.
205-213.
10. Table of limitations and improvements

Limitations Improvements (with algorithm/technique) if


mentioned in paper

1 Activation Function 1 Changing activation function to softmax/relu/sigmoid


for better performance

2 Performance of layers 2 Improving performance of neural networks by


increasing/decreasing hidden layers

3 Accuracy of predictions 3 Using BPNN(back propagation neural network) to


increase the accuracy of predictions

4 4

You might also like