Untitled

I have managed to get the ode running and working!

I tried implementing the new ways they are suggesting but it did not work as expected.

Firstly, I imported all packages that were needed and then declared the sizes of the generations, the population size, maximum depth of the tree the crossover and mutation rates, and the tournament size. I also declared the function and terminal set. I gave a target function which is the termination criteria for the population of programs when it comes to final result. Using this function I created the dataset used when calculating the fitness. I have a class which is used to create the population of programs and represent them as binary trees. This is where I perform crossover and mutation over the trees. Outside of the class I perform the fitness functions and the tournament selection. 

I have started to implement my code into eclipse. I wanted to create this version for sorting a unsorted list of integers, however my first attempt at making my code work is solving a mathematical problem x^2+x+1. 

5. I cannot find the links of a few pdf files I have downloaded onto my device.

I still can’t figure out how to make the fitness function work...

I am now trying to implement the quick-sort algorithm into a genetic algorithm, which is more difficult that I thought. I managed to make a ‘population’ function which creates a population the same size as the genome list size. I then started developing a fitness function which should calculate the fitness cost of each genome, however this is where I have been stuck for a while. I tried coding it in a few ways but i seem to be wrong each time. Unfortunately, I will need some time to figure this stage out. I am unhappy that this is stopping me of completing my milestone as if I could make the fitness function work then I would’ve been able to move on to the crossover and mutation by now. I will be doing more research to find a way to make it work. 

I have managed to get the code to output a random list of five integers. I then started to implement quick-sort algorithm because it is the most efficient sorting algorithm. I had to select a pivot number which separated the data to smaller on the left and greater on the right. I created two functions which formed the quick-sort algorithm. A ‘partition’ function which takes the pivot and separates the data and a ‘quickSort’ function which sorts the integers. When run, the code works fine, as the output gives the random unsorted unsorted list generated by the ‘randomisedList’ and then it also shows the sorted list by the quick-sort algorithm. I am happy this has worked, however now I need to implement the code into a genetic algorithm.

I have now started to try and code a genetic algorithm which takes an unsorted list of integers and sorts then using quick-sort sorting algorithm to generate a sorted version of this list of integers using Google Colab. I am trying to use the information and guides I have read and used in my research. I firstly imported all the necessary packages and then created the list which will hold the genomes. I also made a list which generated the random integers in each run.

2. http://index-of.co.uk/Artificial-Intelligence/Genetic%20programming%20Complex%20adaptive%20systems%20-%20Koza%20J.R..pdf - the book by John Koza I was recommended to read. This is the pdf version.

During this meeting me and my supervisor were discussing the research I have made and the apps I found. He advised me that the SketchAdapt app was very complex and it would be almost impossible to do it on my own for such a short period of time. I did not take that as a discouragement, because realistically speaking, it is impossible for me to complete a complex app like this one alone, because there have been 6 people working on it in a period of few years. Also they have implemented complex techniques, which I am unfamiliar with. We had a long conversation and my supervisor wanted me to do some more research in order to chose wether I want my program to be based on neural networks or genetic programming, because I was unsure. He also recommended me read a book by John R. Koza, about Genetic programming. I was told that, in order for me to see how genetic algorithms work, firstly, I need to build a code that takes an unsorted sequence of digits and outputs a sorted sequence of those digits. It would be very helpful for me to learn and I am planning on starting on it very soon after I have done enough research.

12. https://analyticsindiamag.com/5-languages-to-use-for-genetic-programming/ - which programming languages are the best for genetic programming programs.

The first meeting has ended, we went over my proposal and I would say, the idea which was chosen, was the most complex one out of all 5 of them and I did not expect it to be ‘the one’. However, looking over it, I do realise it is complex enough to push my limits, and expand my knowledge on topics I would have not gone over in that much detail otherwise. The project is going to be a self-coding program, or let’s call it an AI program which writes code independently, without human intervention. Basically, the user will give instructions to what he wants to be coded, and the program writes it. Easy! But not so easy! I will have to be using evolutionary computing, genetic programming, genetic algorithms, program synthesis and machine learning. My supervisor gave me the instructions to research more on the topics and have another meeting when I have done so. Im getting back to work and I will update soon.

Welcome! This is my blog, where I will be updating and showing the progress of my work for my final year computing project. I will be sharing any difficulties I come across during the programming and how have I fixed them, that if I have managed to fix them of course. Let’s get started!