Grasshopper

algorithmic modeling for Rhino

I just edited the questions and replaced the attached files below!

--------------------------------------------------------------------------------------------

Hello,

I have simulated my design with wallacei and have got some questions when I analyse.

Question 1

How do I know the monster of which individual and generational number is?
I would know specific # of individual and generation of this monster, so that I can fix this monster.

Question 2

In the parallel coordination plot, I have got two divided colors. Please see the attached files. Does this mean that my phenotype are divided by two groups-the one is fitted, and the other is not fitted to objectives?

Question 3

I have 4 Objectives. But objective #4 is located outside of the pareto front graph. (x axis is Obj. #1, y is #2, and z is #3.) When I analyze this graph, how can I know where the Pareto-front of individuals of objective #4 is located?

Question 4

Does this “repeated fitness values” mean that the certain repetitions happened per each generation”? Otherwise, the repetition happened for total generations?

Question 5

Does it mean that the more repetition of a certain fitness objective mean the more important that fitness objective to optimize overall fitness criteria?

For example, if fitness objective 1 repetition # is 238 and fitness objective 2 repetition # is 15, does this mean that fitness objective 1 is more important to optimize overall fitness values?

Please see the attached pdf / jpeg files in detail.

Views: 61

Attachments:

### Replies to This Discussion

Hi Mary,

Most of the questions you asked here are addressed in our video tutorials on our YouTube channel and the primer. But in any case, I will respond to your questions briefly one by one.

• Q1: These graphs are Standard Deviation graphs and are drawn per generation. In general, the standard deviation graph is a way to understand the variation within a data set. These curves are drawn by two values, the mean and the standard deviation. Thinner the curve is, the population is crowded around the mean value, so it has less variation. Fatter and flatter the curve is, the population is distributed farther from the mean value, and it has more variation. In Wallacei, standard deviation graphs are drawn from red to blue (first generation to the last generation). Also, if these graphs move towards left, it means the mean value of the fitness of the generation is being minimized, or in other words, this objective is being optimized. So, the curve you pointed out in question 1 is not a monster; it is the standard deviation graph for generation 0. According to the image, the problem is being optimized since the later curves (blue ones) are on the left side of the first-generation graphs. So, one more time, the graph you pointed out is not representing any monster in the simulation!

• Q2: Again red is the first generation, and blue is the last generation. PCP in tab1 only shows the first and the last generation. From top to bottom, phenotypes are getting fitter and fitter (optimized). So based on the PCP, your objectives are optimized.

• Q3: If you read the text, it says objective 4 (Smaller is better). It does not mean that objective 4 is outside of anything; it is just a piece of information for you. Since you have more than 3 dimensions (3 objectives), the fourth objective is being represented as size. And this graph is not a Pareto front graph; it is the objective space. You can toggle the Pareto front individuals in this graph by either checking the checkbox of Pareto fronts in the control panel or by drawing the objective space after simulation finishes.

• Q4: This repetition is for the entire simulation. Based on your image, the value of fitness objective 2 has been selected 15 times. (it is the most repeated since you selected rank 0).

• Q5: It only shows that the specific value of your fitness objective (2 in the image) has been selected over and over in the simulation. It is crucial to know why this happened and see the phenotype associated with these repetitions, but by no means, it indicates that one objective is more important than the other objective. In Wallacei multi-objective optimization, all the fitness objectives are being optimized independently from one and each other. These repetitions may happen because of many reasons such as the limitation of the design space of your problem, etc.

Best,

I read the primer. However, I could not find the explanation about "mutation probability". In the Wallacei default setup, it is "1/n". What does "n" mean? Does it mean the number of genes to be used for the evolutionary process?

If so, how can I count it? For example, if I have 4 genes, and if each gene has the range from integer 1 to 4, then total gene number is 4 x 4 = 16? Then, does 1/n means 1/16 that indicates "mutation probability?

I know I can change 1/n number by un-checking it but I would clarify what n means.

Mutation Probability (0.0 to 1.0) - The percentage of mutations taking place in the generation (Deb et. al. recommends the mutation probability to be 1/n, where ‘n’ is the number of variables (sliders) in the design problem. As such, the default value is 1/n).

• View All