Hello,
Does the KNIME work with three-dimensional data matrices (for example, X-Y- time(time series)?
Thanks
Hello,
Does the KNIME work with three-dimensional data matrices (for example, X-Y- time(time series)?
Thanks
Hi there!
Welcome to KNIME community!
Not 100% sure but don’t think so although you can try to explain a bit more what are you trying to do with your data (visualization, analysis,…) and maybe there is a way in KNIME to achieve it. Also KNIME has MATLAB integration if that is possibility for you.
Br,
Ivan
Thanks for the answer!
Maybe I did not correctly set the topic. There is an idea to apply the KNIME to classify seismic data (for example, lithology prediction, etc., based on seismic attribute data). The learning procedure is clear - we will use a one-dimensional array at the point of the well with the coordinate depth (or time). But how to organize the application of the learning result for a three-dimensional array which is a seismic data cube. The procedure is applied sequentially to each track of the cube - each track has coordinates x-y-time (depth).X and Y constant for each individual trace (track), only the coordinate time (depth) changes
best regards
Anatoliy
Hi Anatoliy @Ant!
Sry for delayed response. Don’t have any experience with seismic data but are you talking here about images? Also is there some specific Machine Learning approach/algorithm you are searching for?
Br,
Ivan
Thanks for the answer. I have a lot of experience in using ML to work with seismic data, but this applies only to some standard methods that are implemented in specialized software products for working with seismic information. I just want to find out the organization of incoming data in KNIME for the application of one or another “classification tree”, which I decide to use. For standard data, which are described as separate cases, a row - no questions. The question is how to organize a filter to separate the rows into separate parts.
Hi there!
A bit clearer now but an example with input data and logic you need to perform would help cause I’m not sure what are you aimng for
Br,
Ivan
@Ant Do you mean that you have multiple measures for each XY coordinate across time (T1, T2, … Tx)? If so, simply using the Unpivot node would do the trick!
Ok, I’ll try. thanks
OK, and if I have the following format - I can transform a seismic cube in the form - xy one column and has for example ten columns of attributes, the eleventh column - the next xy trace and ten columns of attributes, etc. Or do I need to record everything in the form of only rows in series?
Without looking at your data and spending time, I can’t really weigh in. That being said, any n-dimensional array can be represented by Value-Dimension1-Dimension2-DimensionN. From there, it’s just up to you to aggregate or model to meaningful dimensions.