Description
------------------------
Intelligent and automated analysis, interpretation, information mining and prediction based on the huge amounts of data generated by modern biological systems is of crucial importance in research domains such as Bioinformatics, Chemoinformatics and Systems Biology. This advanced level research oriented course focuses on the understanding of computational intelligence and machine learning techniques in Bioinformatics. It covers the latest research topics related to the theory, algorithmic aspects, design considerations and practical applications of machine intelligence techniques in Bioinformatics. In particular, this course is motivated by the potential of applying machine intelligence techniques to heterogeneous biological data generated through next generation sequencing, genome assembly, proteomics and interactomics.
Aims
------
– Concepts of Machine Learning
– Application of machine learning in Biomedical informatics
– Experiment design strategies for machine learning
– Development of required research techniques for a career in MLIB
– Research Output: Projects / Papers
General Information
http://faculty.pieas.edu.pk/fayyaz/
Announcements
| Name | Assignment | Assignment | Sessional | Assignment | Assignment | Assignment | Assignment | Assignment | |
| Differentiation | XOR | Exam | Theoretical | ROC/PR | QP | Backprop | Presentation | TOTAL | |
| Total | 100 | 100 | 40 | 100 | 100 | 100 | 100 | 100 | |
| Weight | 3 | 2 | 25 | 10 | 3 | 2 | 3 | 2 | 50 |
| Dawood | 100 | 80 | 33 | 100 | 100 | 100 | 100 | 100 | 45 |
| Sadaf | 90 | 100 | 34 | 100 | 90 | 100 | 100 | 100 | 46 |
The teaching staff has posted a new lecture notes resource.
Title: Probabilistic Soft Logic
http://www.piazza.com/class_profile/get_resource/jcoztzkoa973d2/jhmnxr5zg223ad
Lecture date: May 30, 2018
You can view it on the course page: https://piazza.com/pieas.edu.pk/spring2018/cis622/resources
What are the 4 major limitations of machine learning?
What are the different types of supervision for machine learning?
What is the main idea behind resnets?
Why would resnets not have the optimization failure which occurs in other neural networks?
Why would the computational complexity of a 152 layer resnet be lower than VGG16/19?
In the next class, please present the basic idea of the papers assigned to you.
Here are the challenge problems for the course:
1. Implementation of Generalized Distillation using pyTorch
2. Extension of Generalized Distillation to arbitrary machine learning problems (regression - must, ranking - if possible)
Solution to the challenge problems will result in a reward in the finals.
This is a good listen...
The teaching staff has posted a new lecture notes resource.
Title: Weak Supervision Notes
http://www.piazza.com/class_profile/get_resource/jcoztzkoa973d2/jhbobfau90m321
Lecture date: May 18, 2018
You can view it on the course page: https://piazza.com/pieas.edu.pk/spring2018/cis622/resources
In yesterday's class, we had discussed how can we classify data without using any training labels. My simple (for demo only) implementation of such a machine learning method that penalizes margin violations and forces a classification boundary throughan area of low data density is given here (https://github.com/foxtrotmike/usvm). As you can see, it does a pretty good job on classifying data without using labels in training.
The teaching staff has posted a new lecture notes resource.
Title: Weak Supervision
https://dawn.cs.stanford.edu/2017/07/16/weak-supervision/
Lecture date: May 16, 2018
You can view it on the course page: https://piazza.com/pieas.edu.pk/spring2018/cis622/resources
| Name | Office Hours | |
|---|---|---|
Dr. Fayyaz ul Amir Afsar Minhas | When? Where? | |
Nauman Shamim | When? Where? |