How can I import a Cell - Array filled with tables of data to the Classification Learner App?

Hello,

I want to use the Classification Learner App to train a model, that can cluster a pressure map from a hand into the different parts of the hand. Like in this picture.

For training data, I already have 400 frames like the one above clustered and for each a table with X and Y coordinate and one column with the cluster, the point belongs to.

I need the model to look at every frame seperate, and not for example at all points from Cluster "5" from all the frames. I believe if I put all of the data in one table, it will cluster all of them at once, instead of frame per frame.

How can I do that? Is it even possible with the app?

Answer 1

[Kartik]

Hi,

Yes, you can train a model using the Classification Learner app in MATLAB to cluster your pressure map data into different parts of the hand. To ensure that the model looks at each frame separately, you can create a separate table for each frame and train a separate model for each table.

[Ruben]

Hi, thank you for answering.

I managed to extract the model as a function and train it 400 times with the 400 different frames using a for loop.

The problem with this and your solution is: I have 400 different models, that havent learned from each other. Thus they are all just as good as a model with number of training data = 1. Or am i getting your idea wrong?

[Kartik]

Can you provide more information about what you're trying to achieve with the Classification Learner App, and what are the input vectors, on the basis of which your model would work?

It is a bit confusing at the moment as making the model learn from all the frames but yet look at each frame seperately, are two requirements which contradict each other.

[Ruben]

Yes sure, sorry.

Right now my data consists of a Cell with 400 entries. Each entry is a 45x3 double matrice, containig a column for X-coordinate, Y-coordinate and the third column with the response variable, the cluster the x-y point belongs to.

I would like to train a model, that learns how to cluster these points (handprint) by going through one matrice after another.

The Classification Learner can´t accept the data type cell though, and I haven´t found another way, to hand over the data without mixing all the matrices.

I hope the problem is more clear to you now.

[Kartik]

The clustering pattern from all the individual frames would be different, for eg.- for the same x and y, two frames can have different clusters. So if we train the model with two same datapoints having different clusters, the learnings of the model might be unpredictable or would favour that cluster more for which there are more datapoints for that particular x and y. Regarding your query for observing each frame seperately, as per my understanding, if we combine the table in such a manner that all the datapoints from every frame are sequentially together, the model would learn from all the points of one frame together.

That means copying the elements of all frames sequentially into the new single table , which would lead to it analyzing frames one after other.

Let me know if it solves your issue or if my understanding of the question is incorrect.

[Ruben]

Yes I now understand how Classifiaction Learner App operates and what I need to do. Thanks for the good support!!