Science topic

# Self-Organizing Maps - Science topic

Explore the latest questions and answers in Self-Organizing Maps, and find Self-Organizing Maps experts.

Questions related to Self-Organizing Maps

Recently, I discovered the dimension of the SOM network do turn out to be the number of data clusters for data clustering or image segments when used for image segmentation.

For example, if the dimension of the SOM is 7 x 7, then the number of clusters(segments) would be 49, if the dimension of 2 x 1, then the number of clusters(segments) would be 2.

1. Therefore, are there techniques for determining the dimension?

2. What should be the basis/yard stick for picking the dimension?

3. If the knowledge of the data is the basis/yard stick for picking the dimension, is that not a version of K-means??

There are different beneficiary effect of soil organic matter in the soil. Therefore, you have to mention at least seven beneficiary effects of soil organic matter (SOM) in the Globe.

So I am analyzing my soil data. Interestingly I have found that there is an inverse relationship between CEC and clay content (%). In many studies, I have seen this relationship is positive. However, some studies highlight that this may be the case if the soils are sandy, very acidic and have low soil organic matter. Which means that my soils get their CEC from SOM that clay fractions. Any ideas on how I can explain this.

Hello everyone,

I have set of design variables corresponding to Pareto optimal solutions. Now, to define the correlation between design variables and objective functions, some researchers used SOM technique. The clustering example using SOM is attached herewith.

I have to do the same clustering for my data, as shown in attached image.

Can anyone suggest which software MATLAB/Python/any other is better and any useful tutorial links for implementation for the same.

Hello all, I have an experiment in which I train a 5x5 map of nodes to describe meteorological fields with Self Organizing Maps. I am using the python library called minisom but I've tested a few others and have the same issue. I can initialize my som model by performing a principal component analysis and distributing my initial nodes along the main axes, but I think the library does not allow me to control the node values in finer detail. For example, if I train a som model with certain parameters, I obtain a set of nodes, and then I want to use those trained nodes to initialize a new training experiment, I can't do that, because the python "som" object only allows me to "see" the node values, not to change them or specify them in any other way than training.

So, I would like to know if I can achieve this with these simple python libraries or if I should be looking for more complex tools to train som models.

Do you still trust the conventional factor of multiplying SOC by 1.724 to obtain SOM? Or is it not better to report in SOC only?

Is there a use of machine learning algorithm (e.g. Self-Organizing Maps (SOM), Graph Convolution Neural Networks (GCNN)) for vector spatial data using topographic maps? e.g. seperating/clustering linear features (transportation or hydrography) such as

- highways vs. railroads
- ditches vs main rivers

dear researchers

I would be appreciative if you let me know your opinion about the disadvantages of the SOM clustering algorithm.

Please guide me

I group the data in 2D-SOM (Self Organized Map) method. SOM was used with a 25 x 25 layer with a rectangular grid topology. The data was 717 in total. I have grouped out a total of 625 groups.

Groups which exist at 1 row = 307

Groups which exist at 2 rows = 218

Groups which exist at 3 rows = 120

Groups which exist at 4 rows = 56

Groups which exist at 5 rows = 10

Groups which exist at 6 rows = 6

Based on the information above, how should I split data into training/validating/testing sets. So that the mean and the standard deviation are similar. Please help advise

Please suggest full form for SOM in reference to data mining, machine learning and data science except Self organising maps. i have used the abbreviation SOM in my topic of research that stands for self organising map.But now as my work contain many other data mining algorithms also, so i need a broad perspective. Moreover I cannot change title of research.

I grouped sites in six different clusters on the basis of fish species abundance using a self-organizing map (SOM). Now, I want to determine to what extent water chemistry variables can explain this grouping. However, I don't want to assess the linear relationship between the fish communities and water chemistry variables.

Since SOM is not a linear clustering method, can I use the LDA?

Would the same answer apply if I had used Cluster Analysis for grouping?

Hamed

Could you please help me and give some tips on how to design trajectory/line using a one-dimensional Kohonen "map"? I need it to my article on bankruptcy but as I'm not statistician I can't handle this. More spcecifically I don't know how to modify the R script for two-dimensional map to have one-dimensional.

I've some question about clustering data set using SOM. What the pre requisite of the data?

I'v mixed data types (categorical, binary, continues).

I'transformed categorical to dummies then merged my data set into one. I did not apply min-max scaler as I've many dummies variables.

I preferred using pca. I applied pca to reduce features by feature importance. then I use these axes as input in my SOM (SOMPY).

Is this a good approach?

If my binary data has for example 95% of 1 and 5% of 0 or vice versa, is this a problem for clustering?

Clustering - Neural Netowrks

what is the best software to generate SOM and which algorithm could be adequately used in this type of researches.

thank you all.

Using Self Organizing Maps algorithm to cluster some data will give us NXM centroids where N and M are pre-defined map dimensions. If I have a distribution of species and want for example to get 4 centroids. I am not sure it is correct to choose for example 2x2 SOM dimensions.

Thus, I put 10x10 map then use the K-means to get 4 centroids. However, if I use 8x8 map I will get different centers. So, my question is what is the best NxM dimension?

Dear all;

I am using MATLAB and K-means to group online students with similar interaction on e-learning. But how can I determine a good number of clusters?

Thank you.

There are different types of network creation functions which can be employed in artificial neural network such as cascade-forward network, competitive neural layer, distributed delay network, elman network, feed-forward network, function fitting network, layered recurrent network, linear neural layer, learning vector quantization (LVQ) network, nonlinear auto-associative time-series network, radial basis network and self-organizing map.

One of the most common networks is feed-forward backpropagation neural network. Is it better than the other networks in any aspect? If so, what are the advantages? And does this network have any noticeable drawback?

Hello, everyone.

Someone knows if it is possible, and how to do, to work with NaN data using the "selforgmap" function in Matlab? I have some variables that have fails in the data, so I would like to test the SOM's ability of overcome this problem.

Hello

Dear all

These question from petrel,

1. How to build a new and desired attribute in the petrel?

2. Can I export of cube seismic and run in MATLAB? How?

3. Is there a place to give a number in color adjustment? For example, tell that the amilitudes with value 2, give yellow color ?

4. Can neural networks and self-organizing maps and fuzzy logic be done with petrel?

Teuvo Kohonen self organizing maps comes under unsupervised learning in machine learning. One of the most advanced techniques of all time. Will be very useful for data analysis.

Need to know self organizing map and its tools (available packages)

There are many plug-ins for the Weka SW on the net, but I haven't found one that implements the SOMs. Have any of you implemented it?

Even a Beta version can be fine.

Thank you very much.

I have never used a modeling software and now I find myself in a research work that has to do with environmental modelling. Will using SOM be a good option for me in terms of mapping out infectious disease clusters (spatial temporal modelling)? Is it easy to learn how to use it? How can I get the software?

From the documentation of SOM, I understand that we use SOM to visualize the underlying pattern in the higher dimensional data. However, since SOM is basically a neural network, so do we need to supply higher dimensional discriminative features extracted from data or the rough data directly and SOM will extract discriminative features automatically. An example of rough data could be the time domain or frequency domain responses of the pristine and defective mechanical structures.

Neighbour function in SOM is one of the network components.

Is it possible to improve network performance by implementing unused until now neighbour functions?

For example, in MLP even sine function is proposed for neuron activation.

For those who are experts in the field of clustering I am looking at the possibility of taking high dimensional data such as faces to reduce the dimensionality of the data. then use Affinity Propagation to collect what could be a "true" training set. these are faces that best represent the entire data.

Currently with the Olivetti faces the dimensionality is already reduced by the faces being black and white, very little background, etc. I would like to use SOM on real images then collecting a simiarlity matrix based on these findings in a lower dimension, use it in Affinity Propagation to collect a "training set". My goal is to see if I can improve the training of an SOM on real data by using Affinity Propagation to give us optimal training sets.

I'm learning self-organizing maps, however I don't know how to determine the number of nodes by which the data will be well classified. Which metric may be useful?

Hello，I wanna know will this project quantify the process of trophic flow in reef ecosystem? Since it may be difficult to characterize the trophic flow form inorganic sectors( e.g.nutrition) to organic sectors(e.g. SOM,POM,PP ).

Hello,

Recently I finished analyzing a series of soil samples in order to determiner the carbon content of the samples. The samples were collected at different depths and vary in carbon content, from highly mineral (coarse sand) to organic (peat and mixture of sand and peat).

First, I did the total carbon (TC). To do so, I sent my samples to a lab, which used a LECO628 Carbon/Hydrogen/Nitrogen determinator to give my the CHN concentrations. Beforehand, I had dried and crushed all samples.

After, I used the loss on ignition (LOI) method to determine the fraction of organic carbon. The samples were heated at 475°C during 6h, then the oven was turned off and the samples were left in the oven overnight. Then, I converted the soil organic matter (SOM) to soil organic carbon (SOC) using a 58% correction factor (since SOM has conventionally been assumed to contain, on average, 58% SOC).

However, to my great surprise, all SOC values are ~22% higher than the TC values, which does not make any sense. I verified with the lab about the TC results and everything seemed normal when done (the analyses were done in two batches, so if there had be a mistake with the blanks it would have shown in half the samples). I also wondered about the 58%, thinking that the value does not fit my samples. However, the relationship between SOC and TC is consistently linear throughout all carbon concentrations, indicating a systematic error (see attached file).

Has anyone ever seen or heard about a similar problem? Any suggestions about the source of the problem or where I could have made a mistake?

I have 8 different patterns made up of trangle, rectangle and different combinations of trangle, rectangle and sine waves. I have large number of training and test data sets. I want to train a SOM for classifying the data sets into 8 groups. The size of each of our pattern is around 1000. i.e. the vector size of the pattern is 1000. Since I have large no. of patterns (around 2500), it takes very long for SOM to train. How can I use PCA to reduce dimentionality and hence accelerate the training? At present, the accuracy of classification is very poor. Is there any other method better suited for this problem?

Hi everyone. I'm currently trying to assess the ability of human volunteers to cluster a set of images into a fixed number of clusters according to perceived visual similarity (the images are self-organized maps of breast tumours' gene expression). In order to do that I build an averaged identity matrix (with binarized similarity: 1 is belonging to the same group, and 0 not belonging), which I later transform into a graph and perform the partition. I'm stuck in the part in which I contrast human ability to an algorithm. I was wondering if there's an algorithm or software which transforms a set of images into an undirected wighted graph, where the weight of the edges represents the similarity between images (nodes). This would be my best scenario, because I would be able to perform the same partition method and compare. Else, I'd still appreciate some suggestions for image clustering according to visual similarity. Thanks in advance.

In Geo-science there has been several publications on Self organizing maps, Support vector machine.

I am doing a research to compare the effect of 3 land uses ( forest, irrigated crops and rainfed cereal crops) on available phosphorus (AP) and N_NO3 of alfisols and mollisols. for AV i found that the highest value was under irrigated crops and the lowest was under rainfed land use . it can attributed to fertilizers application in irrigated crops while no add fertilizers in other land uses . i asking what the reason for decreasing AP in rainfed land use comparing to forest land since both no fertilizers ere add. can we attribute that to difference in SOM which was higher in forest land .

2- is it logical that there no difference between rainfed and forest land in their content of N-NO3 .

SOM increases available water capacity, also AWC increases with increasingly fine Stextured soil, because fine textured soils have a greater occurrence of small pores that hold water against free drainage. If there is land converted from forest to cultivation, which led to significantly decrease in SOM and increase in clay% and bulk density. what is expected to happen to field capacity, wilting point and available water? both lands ( before and after conversion) were clay texture .

Dear researchers,

I am using SOM to cluster my data in python 3.6 and I have get the result visually through various maps. My question is how to obtain the results (description) of clustering process such as: members of each output cluster, the distances between these clusters? This is required to check the coherence factor inside each cluster and the level of dissimilarity between the overall clusters.

Thanks

Dear researchers,

I am using SOM to cluster my data in python 3.6 and I have get the result visually through various maps. My question is how to obtain the results (description) of clustering process such as: members of each output cluster, the distances between these clusters? This is required to check the coherence factor inside each cluster and the level of dissimilarity between the overall clusters.

Thanks

how to show the relation between SOM and Precipitation Mainly based on point scale data?

Dear all;

I used SOM and K-Means clustering methods in MATLAB to cluster my data. After the clustering, is there any function that I can use to test which of them has the higher performance since there is no prior knowledge about the data?

Thank you.

I am working on network fault clustering using Fuzzy C-Means and I would so happy anyone could help me if there is better method than Subtractive Clustering to find cluster centers and provide the optimal number of clusters?

What is the difference between partition around medoids (PAM) and density based clustering algorithm (DBSCAN)?

I want to use Self organizing map technique of unsupervised learning to classify supernova 1a. How to use spectrum and light curves of Supernova 1a with Self organizing maps (SOM).

I want dataset for parent child hierarchy for event sub-event detection. Can anyone point to me to the link

I want to extract patterns from suspended sediment concentration(SST) imagery using self organizing maps in matlab,according to the paper of Richardson et al.(2003) "Using self-organizing maps to identify patterns in satellite

imagery" .

Now, I want to know how to pre-processing my SST data and how to perform the result of SOM as an image.

Thank for your help.

I have 140 samples data about driver driving , and want to divided them into specific number of types,such as three .

I have already try to do that ,use a 1X3 map (SOM) in matlab nctool to divide them into 3 types , and consider that each neuron represent one center of the type.

Does it is proper to do that ?

Thanks for attention !

Recently, i have downoald climatic data from Swat database, but i need the name of climatic station which served to built this database. I have som coordinate but i hav'nt the name of stations.

Can some boby help me for this work?

thanks for all.

I used SOM in Matlab to cluster data and I knew how to get the cluster centers. How can I obtain the indices of my input vectors that are around each neuro (cluster center)?

I am seeking for a reference paper which utilized SOM as a basis method for clustering and use a small sate set (i.e. data collected via survey with low rate of response?)

I want to use SOM and Maximum Dissimilarity Algorithm (MDA) to split Input dataset to three subsets ,i.e., Training, Testing and validation subsets for regression problems. To use these subsets in Data-driven models to train and validate and then test the model based on the selection method Of SOM and MDA.

Thanks in Advance

Is there any stopping criteria that i should use to make the algorithm converge rather number of iterations??

how can i initialize SOM nodes' weights,instead of random initialization?

Which is the proper way to validate self organizing maps (SOM)?

Dear all

When i learn SOM ,all training set are used each iteration ?or just one training vector for each iteration??

Hi all ,how to use SOM for multi-dimensional images if i don't have a ground truth table ?what should be entered to SOM?the whole image or random subset?? and how to evaluate the result??

i want to make Self organizing map for upregulated and downregulated genes under salt stress at four different time point....

I want to implement a set of self-organizing maps that represent various features of the dataset via different distance metrics, and ideally can also do hierarchical clustering on top.

I'm using the SOM tool in SAS enterprise miner to classify a dataset of 666 records into clusters. If I change the order of the observations, the SOM algoritm produces completely different clusters. I can't understand if it is a problem of convergence or if I did anything wrong in setting the options.

Did anyone get similar problems?

I am using kdd data set. And want to label all the features using a umatrix or is there any other option to do that. Is there any function in matlab to do that?

Is it possible to quatify a sammon's projection, i.e. transform it from a projection into a measurement?

After using the Self Organising Map (SOM) function in Matlab to cluster a set of high dimensional data, I have had to write additional lines of code to obtain the centres of the clusters, which introduces extra computation to my Matlab code. based on my understanding of how SOM works, I think there should be a matlab command that should reveal these centres, but i cant find it.

1. Mathematical definition of the SOM.

2. How to develop SOM in MATLAB.

Hi all, I am attempting to measure magnification (see for instance www.mitpressjournals.org/doi/abs/10.1162/089976606775093918) in practical experiments with neural gas applied to geometric inference.

I found a clear description of how this measurement can be done in practice in http://www.researchgate.net/publication/6307269_Explicit_magnification_control_of_self-organizing_maps_for_forbidden_data/file/9fcfd5086521b0d268.pdf (see appendix).

Does anyone have any suggestions or comments about the measurement method?

i need to improve the results from self organizing map by using local binary pattern

Now I want to apply nighttime light imagery and gridded point of interest(POI) to classify the function area in Beijing. I used SOM (self-organizing map) and got some results of 4X4 class (by my decision), but how can i make sense of it? Is there any classification standard to refer to?

I'm looking for a method for unsupervised classification of big data with an unknown number of clusters. Can you suggest a robust method? Is there any Matlab toolbox dedicated to this purpose?

I hope to find some recent works on LVQ a.k.a supervised SOM networks and successful applications.

I want to modify SOM NN for design CPNN algorithm, face recognition system using CPNN algorithm.

I am looking of the new trend SOM researches. Up to your knowledge, would you suggest some sources or upcoming ideas for SOM development?

I have calculated weights and input data vector but have problem to calculate y

I am interested to work on optimization of SOM using Ga algorithm

The aim is that each neuron (or node) of a SOM (self-organising map or Kohonen map) gets assigned roughly the same number of input vectors.

I am trying to combine both methods for classification of human grasping.

I know Multidimensional Scaling uses only a distance matrix, but Self-Organizing Map requires coordinates of points in the original space. What are some other dimensionality reduction techniques, such as Multidimensional Scaling, that need only a distance matrix rather than point coordinates?