I remember I read a research paper a few years ago in which the authors had employed a genetic algorithm for identifying bugs in a piece of software. The algorithm works by generating a large population of individuals. Each one of them is a valid input to the program. By valid I mean something that the program can accept as input. So the search space is the set of all the valid inputs to the program at hand.

The algorithm evaluates all the individuals by seeing how the program works as they are input to it. On the inputs on which the program crashes, the algorithm uses them to evolve further generations. This is how the algorithm narrows down on the search space and moves to the regions which generate inputs that crash the program. On further inspection of the inputs, the user may learn about the patches of the code of the program that may have caused crashes. The bugs can then be fixed easily.

The talking I am talking about was quite old. I guess it was written in the late 90s’ by a few researchers in the Norway’s Trondheim University of Technology.

The scheme was elegant nonetheless. It reduces a lot of labor on part of the software engineer as the process of bug testing is automated.

Search based software engineering is a matured field now. Every now and then I get an email about a call for papers from a conference, symposium or a workshop being held on evolutionary computation and there is almost always a track for search based software engineering.

Here I am posting a few links that might be of interest to anyone who is willing to work in this area.

Here is a link on GECCO’s website about its track on search based software engineering.

GECCO 2017 | Program Tracks

This is a link about an independent symposium that is held about this topic.

http://ssbse.org/2016/

Evostar conferences are leading European conferences about various topics in theory and applications of machine learning. Here is a link to their conference of software engineering.

http://www.evostar.org/2017/cfp_evoset.php

I shall keep on finding and posting more valuable stuff on this topic here.

Photo by ikewinski

I was stunned by his question that how long I was taking to prepare my lectures. My reply was that I had nearly spent five weeks to prepare my lecture on normalization (although I have spent more than that). The reason I spent so much time in preparing for normalization is that I had eyed it in the beginning of the course as the most important topic in databases. And, indeed, it is.

Anyhow, the reason he surmised that I was not preparing my lectures well was grounded in his observation about the way I taught about functional dependencies. A student asked that should X always be a key attribute in a functional dependency X->Y? My answer was no, it need not be. I am still adamant about my answer. So in a functional dependency such as X->Y, either one of X or Y can be a key or non-key attribute. The idea behind normal forms is to address issues related to key and non-key attributes.

A succinct definition of a functional dependency, such as X->Y, is that whenever X assumes a certain value (say x), and for that value of X (x), Y assumes a certain value y, then in every subsequent retrievals of X, whenever X assumes the value x, Y should assume the value y. Thats it! Functional dependency is no more than that. It is a very simple mathematical idea. It is the job of the subsequent normal forms to reduce the dependency of non-key attributes on non-key attributes.

I had to haggle with my friend a little bit to convince him about my point. And that discourse kept me thinking about functional dependency for quite some time. One of the questions I pondered about was how to figure out a functional dependency between attributes, which are nothing more than variables in a mathematical sense.

A very easy way to figure out functional dependencies between attributes is to review the cardinality of their relationships. Just keep in mind the definition of functional dependency I gave in the paragraph above (actually two paragraphs above). Now if two attributes (X and Y) have a zero-to-zero relationship, what is the functional dependency going to be? They shall have no functional dependency! They shall be independent of each other as nothing in X maps on Y.

Similarly, in case of a one-to-one relationship between X and Y, there will be a bidirectional dependency meaning X->Y and Y->X shall hold.

In case of a one-to-many relationship between X and Y, where X is on the one side and Y on the many side, the functional dependency Y->X shall hold. The reason is that many values of Y shall map on a single value of X, so Y-X. But one value of X shall map on many values of Y, so X->Y shall not hold. Similarly, for the converse, i.e. many-to-one relationship between X and Y, the functional dependency X->Y shall hold.

For the many-to-many relationship between X and Y, there will again be no functional dependency between the two. The reason is that a value of X shall map on multiple values of Y and vice versa, thus violating the definition of functional dependency.

What do you think would be the functional dependency in the case of a zero-or-one-to-many relationship? I leave it for you guys to figure out.

Photo by mattacevedo

A nice and succinct tutorial on ERDs is as follows.

What is entity relationship diagram? – Definition from WhatIs.com

A Quick Guide to Database Normalizing with First Normal Form (1NF)

Valuable tutorials about normalization.

Normalization in DBMS: 1NF, 2NF, 3NF and BCNF in Database

Normalization is a process of organizing the data in database to avoid data redundancy, insertion anomaly, update anomaly & deletion anomaly. Let’s

1NF, 2NF, 3NF and BCNF in Database Normalization | DBMS Tutorial | Studytonight

Here is a very nice course about database systems by MIT OCW.

Database Systems

Here is a nice course on Database management by Coursera.

Database Management Essentials – University of Colorado System | Coursera

A nice tutorial on 4NF.

Database Normalization: 4th Normal Form

Beyond Boyce-Codd normal form with 4th normal form

A nice tutorial on 5NF.

Database Normalization: 5th Normal Form and Beyond

Normal forms beyond 4th are mainly of academic interest

Fifth Normal Form (5NF) – 5NF in DBMS – 5NF in Database

Fifth Normal Form – A database is said to be in 5NF, if and only if it is in 4NF anlong with additional conditions which are explained here.

Photo by benjami