Solve grasshopper geometry problems using ML

Hello
First, Sorry if I am not asking the question in a proper way!
I am trying to discover some kind of problems/bugs- maybe related to data structure - or geometrical ones that face users in grasshopper and can be solved applying machine learning algorithms.
I know there are some plugins do clustering and so , but I mean which kind of problems can I start thinking of ?
Thanks

I’m not sure what you intend to do, but it looks like you’re starting the house from the roof down. What do you mean by

and

and

The first thing you need is access to the data that represents the problem to be solved. Either with a data set to be able to use supervised learning or with a program that is able to recognize the data you are looking for, to use unsupervised learning. Do you have experience doing programs inside the guts of Grasshopper?

Hello Dani
Thank you so much for your reply!
I mean for example clustering algorithms like k-means / affinity propagation can solve problems related to clustering mesh faces for fabrication processes. so I am looking for other kind problems that could be solved using ML algorithms.
Yes I have some modest experience, I tried before supervised learning with external data set from UCI using grasshopper python and python remote to import libraries like sklearn and pandas.

Rhino has methods for approximating a line or plane to a point cloud. These algorithms are also used in ML. But that doesn’t make them ML, they’re just algorithms used in ML. They’re different practical interpretations of optimization techniques.

Yes I understand that machine learning process from pre-processing till evaluation of prediction is different than just using statistics algorithms.
That is why I am asking about kind of problems can be solved by ML as a process, like clustering elements or spaces of architectural plans , will require a set of maybe 1000 or more plans before predicting new one. and I know it is not always easy to find data sets suitable for training, so my question regarding also problems that I can find available data sets for.

My point is that optimizing geometry is not an ML problem, it is an optimization problem, if you don’t need to train a model to solve it. For prediction you need a statistical model, like ML algorithms. So your example is not an ML example, just another example of how confusing language can be if you are not rigorous.

Forget that, going back to your question, if you want to use machine learning, look for problems that can be solved by learning using statistical patterns, or that can be represented or evaluated (in general, measured) in another more interesting way.

No I didn’t mean optimizing geometry at all, but I mentioned that it needs a dataset for training.
Maybe I asked in an inaccurate way, my mistake.
Thank you !

I agree here with Dani. Whatever you are trying to do with ML make sure you understand well enough different ML algorithms to be able to choose which one is the best one for your goal, or if one is really needed or not for your problem. For example RL does not need any training set. Another very important thing is to make sure you know at least the required mathematics such as Calculus, Linear Algebra, Statistics and Probability so you can understand what the heck is happening. After you are technically agile, I guess you would also need to be to be good enough to know which libraries to glue together to make things work.

https://developers.google.com/machine-learning/crash-course/prereqs-and-prework

Mastery of intro-level algebra. You should be comfortable with variables and coefficients, linear equations, graphs of functions, and histograms. (Familiarity with more advanced math concepts such as logarithms and derivatives is helpful, but not required.)

In other words… what they really mean, in my opinion, but dont say it to not scare people away, is the more math you know, the more successful you will be to understand the algorithms and thus, knowing how each one works and when to apply any particular algorithm.

PS. Machine Learning math is %$%@$$!£ hard… even for Computer Science mayors, which have hard core mathematics in their careers. Imagine what it took to program Alpha Go… surely much more than intro-level algebra…

1 Like

Hello Nicholas
Thank you for your reply!
Yes I agree with you, I have attended one course on Edx for needed math for ML more than one year ago but yes I need of course a refreshment.
Thank you very much for sharing this link, it is very helpful.

That is great, then you are a little more than halfway there then

1 Like