i have almost no experience applying ML to practical problem in real engineering world, my observations are limited to some examples i have seen in different courses.
I need to solve a optimization problem so that and maybe later get a model that generalize.
The goal is: to maximize the output based on three of the inputs, in another word: What should be values of these three input , to get the max output
It seems to me a way of modeling my function and inputs/outputs is a non linear regression, but how i can solve the optimization , i have no idea
I would appreciate your advices , how should i approach this problem?
What should i define the inputs that i need to tune? should i use them also as label?
I have some other question but first i need a start point
Thank You ver much in advance
You should be starting first doing some courses in here, I dont think we could explain you all the information included in the courses. You may start with the Machine Learning or Deep Learning specialisation if you have some python coding experience.
i have good experience in c programming and some experience in python, i already visited some ML courses for beginners and i am currently in third course of DL spezialisation…
the problem is all examples are vision, tabular excel data etc
all examples i have seen in all different courses were made with prepared datasets.
I couldn’t find even one example were the solution would be from scrach, were you should start thinking what should be taken as label and what should be taken as feature…
i am searching this kind of approach.
even refering to a good book would be nice, if anybody know
Hello @Mohammad_Foroughi! The thing that you want to predict is called the target variable, labels, or dependent variable, for example, a disease from an X-ray or a house price based on the number of rooms, areas, etc. The thing on which you want to base your prediction is called the feature, such as the X-ray image or the number of rooms, areas, etc.
Is it answered your question? If not or if you have any other queries, let me know.
I think you’re are searching for Machine Learning Specialization. The questions you have are answered during the Courses in this specialization and they are answered in a very “from the ground up” / “from scratch” approach.
I agree % - there are many orchestrated courses with perfect datasets but there is a lack of courses where the datasets are “dirty”, where the model building is iterative and mistakes are not redacted out, etc. - more like in a real world setting.
But on the other hand these kind of approaches are more:
- domain specific - Regression, Natural Language, Vision, etc,
- expertise level specific - from explaining sum or multiplication (kindergarten math) and basic programming to derivative calculations, Jacobian matrices and algorithm optimizations
- also, maybe no expert wants to expose their “silly” mistakes on the internet or maybe they don’t find it informative (to which I would not agree)
Anyways, I hope you find Machine Learning Specialization as a good framework to dig deeper into the area you find most interesting or useful.
thanks for your answer, my main question was the approach of an optimization problem in Machine learning.
Morever the choice of features , in a dynamic real system we have lots of different data that have direct or indirect impact on target values, how to choose them generally when we don’t have a big dataset yet and we can not test and check and compare their impact on output
thank you, i am more searching to a methodology of how to define a mathematical problem in machine learning …
Now i am middle of the project nad preprocessing my data in Matlab is finished, there are crazy amount of data in my dataset, and maybe i don’t need all of them … i am looking for a model that can generalize and make predict for future dataset
- Do you think starting this course in such short notice would help me?
- this course’s launch year is 2012, is there any information about last update and more detail of this course?
Thank you very much
In my humble opinion it very depends on your experience with these kind of problems and please don’t get me wrong, I have the impression that you were sincere that you “almost have no experience”. In this case this course would be very helpful to cover the basics (in particular - feature selection, feature engineering, scaling and etc.) and even if you are in the middle of something, fixing mistakes now or not making them is always important.
So I cannot say for certain, but I think you would benefit from this course a lot. But on the other hand if the house is on fire… there’s no time to learn
This course was recently updated (1 year ago (in 2022) if I’m not mistaking) and to be honest the fundamentals have not changed. I can assure you that none of the techniques described are obsolete.
Either way, whatever you choose please let us know how you’re doing
well i don’t know if that counts
But my experience is limited to applying samples and datasets from different courses to ML, from retail datasets to medicine to some image processing
But never i applied ML to a real problem with raw data
I almost know all the basics… i will give it a shot