I've got a dilemma on deciding an optimal number of cluster from my dataset.
I have obtained ConsensusClusterPlus heatmap plots but not really sure on deciding on the optimal number of clusters.
Can anyone help in this? I've attached pictures of heatmaps and plots to decide on optimal number of clusters. Seems like k=3 is an optimal number click here to see pics pic 6, pic 4, pic 2, pic 3, pic 7, pic 8, pic 9, pic 10
Note, I'm not that specialized in this topic, but currently I'm also working with cluster ensembles. So, take my arguments with caution.
Based on the images (2:7) the heatmap shows a 2-block structure, giving a slight hint of k=2. Going further with picture 9, the elbow similarly suggests the same k, since after two there is a very sparse decrease in change of the AUC.
In addition, in case you haven't read it, see the paper of Monti, Tamayo, Mesirov and Golub (2003) https://link.springer.com/article/10.1023/A:1023949509487 Using heatmap aswell. However, also take note with the weakness surrounding a heatmap, such as deriving clusters from data while there appears to be none, see Șenbabaoğlu et al. (2014) http://www.nature.com/articles/srep06207.
I assume you already found an answer, so please share your argument if possible, as clustering is often underpinned as a subjective type of class discovery in ML.
Related
I've tried to repeat the same results with the same flow, and I don't understand the results are different in each situation.
I describe the situation I have a file with 192 instances and 37 features, y select in all cases the same columns and preprocess by Median and StdDev. It computes the PCA with 7 principal components. The following step is to run the k-means algorithm (k is between 2 and 8) from this 'new' dataset. The scatter plot shows the results for k=5.
I attached different images with my flows.
Image1: original flow
The first one is the original flow (it is painted of yellow color), which I would like to repeat without the rest of the options (the second image).
Image2: flows repeated
However, when I tried to do it, I saw that the results are different (the third image) Of course the colors don't determine the differences, however the clusters are different. In addition the Slhouette Scores are different too for the different flows.
Image3: results of the different flows
K-means initializes with the kmean++ and I have the question if I can "control" this, or if the way to initialize k-means is always randomly. I saw in other programmes that there is an option called seed which is used to control that an experiment can be repeated but I didn't see this option here or something similar.
I wonder if it is possible to obtain always the same results with the same flow (using k-means).
It seems that the issue happens because no random seed is set in the k-means widget. So initialization is different each time you repeat an experiment and because of nature of your data, the method converges differently. Can you please report your issue to Orange3 issue tracker.
Just starting to play around with Neural Networks for fun after playing with some basic linear regression. I am an English teacher so don't have a math background and trying to read a book on this stuff is way over my head. I thought this would be a better avenue to get some basic questions answered (even though I suspect there is no easy answer). Just looking for some general guidance put in layman's terms. I am using a trial version of an Excel Add-In called NEURO XL. I apologize if these questions are too "elementary."
My first project is related to predicting a student's Verbal score on the SAT based on a number of test scores, GPA, practice exam scores, etc. as well as some qualitative data (gender: M=1, F=0; took SAT prep class: Y=1, N=0; plays varsity sports: Y=1, N=0).
In total, I have 21 variables that I would like to feed into the network, with the output being the actual score (200-800).
I have 9000 records of data spanning many years/students. Here are my questions:
How many records of the 9000 should I use to train the network?
1a. Should I completely randomize the selection of this training data or be more involved and make sure I include a variety of output scores and a wide range of each of the input variables?
If I split the data into an even number, say 9x1000 (or however many) and created a network for each one, then tested the results of each of these 9 on the other 8 sets to see which had the lowest MSE across the samples, would this be a valid way to "choose" the best network if I wanted to predict the scores for my incoming students (not included in this data at all)?
Since the scores on the tests that I am using as inputs vary in scale (some are on 1-100, and others 1-20 for example), should I normalize all of the inputs to their respective z-scores? When is this recommended vs not recommended?
I am predicting the actual score, but in reality, I'm NOT that concerned about the exact score but more of a range. Would my network be more accurate if I grouped the output scores into buckets and then tried to predict this number instead of the actual score?
E.g.
750-800 = 10
700-740 = 9
etc.
Is there any benefit to doing this or should I just go ahead and try to predict the exact score?
What if ALL I cared about was whether or not the score was above or below 600. Would I then just make the output 0(below 600) or 1(above 600)?
5a. I read somewhere that it's not good to use 0 and 1, but instead 0.1 and 0.9 - why is that?
5b. What about -1(below 600), 0(exactly 600), 1(above 600), would this work?
5c. Would the network always output -1, 0, 1 - or would it output fractions that I would then have to roundup or rounddown to finalize the prediction?
Once I have found the "best" network from Question #3, would I then play around with the different parameters (number of epochs, number of neurons in hidden layer, momentum, learning rate, etc.) to optimize this further?
6a. What about the Activation Function? Will Log-sigmoid do the trick or should I try the other options my software has as well (threshold, hyperbolic tangent, zero-based log-sigmoid).
6b. What is the difference between log-sigmoid and zero-based log-sigmoid?
Thanks!
First a little bit of meta content about the question itself (and not about the answers to your questions).
I have to laugh a little that you say 'I apologize if these questions are too "elementary."' and then proceed to ask the single most thorough and well thought out question I've seen as someone's first post on SO.
I wouldn't be too worried that you'll have people looking down their noses at you for asking this stuff.
This is a pretty big question in terms of the depth and range of knowledge required, especially the statistical knowledge needed and familiarity with Neural Networks.
You may want to try breaking this up into several questions distributed across the different StackExchange sites.
Off the top of my head, some of it definitely belongs on the statistics StackExchange, Cross Validated: https://stats.stackexchange.com/
You might also want to try out https://datascience.stackexchange.com/ , a beta site specifically targeting machine learning and related areas.
That said, there is some of this that I think I can help to answer.
Anything I haven't answered is something I don't feel qualified to help you with.
Question 1
How many records of the 9000 should I use to train the network? 1a. Should I completely randomize the selection of this training data or be more involved and make sure I include a variety of output scores and a wide range of each of the input variables?
Randomizing the selection of training data is probably not a good idea.
Keep in mind that truly random data includes clusters.
A random selection of students could happen to consist solely of those who scored above a 30 on the ACT exams, which could potentially result in a bias in your result.
Likewise, if you only select students whose SAT scores were below 700, the classifier you build won't have any capacity to distinguish between a student expected to score 720 and a student expected to score 780 -- they'll look the same to the classifier because it was trained without the relevant information.
You want to ensure a representative sample of your different inputs and your different outputs.
Because you're dealing with input variables that may be correlated, you shouldn't try to do anything too complex in selecting this data, or you could mistakenly introduce another bias in your inputs.
Namely, you don't want to select a training data set that consists largely of outliers.
I would recommend trying to ensure that your inputs cover all possible values for all of the variables you are observing, and all possible results for the output (the SAT scores), without constraining how these requirements are satisfied.
I'm sure there are algorithms out there designed to do exactly this, but I don't know them myself -- possibly a good question in and of itself for Cross Validated.
Question 3
Since the scores on the tests that I am using as inputs vary in scale (some are on 1-100, and others 1-20 for example), should I normalize all of the inputs to their respective z-scores? When is this recommended vs not recommended?
My understanding is that this is not recommended as the input to a Nerual Network, but I may be wrong.
The convergence of the network should handle this for you.
Every node in the network will assign a weight to its inputs, multiply them by their weights, and sum those products as a core part of its computation.
That means that every node in the network is searching for some coefficients for each of their inputs.
To do this, all inputs will be converted to numeric values -- so conditions like gender will be translated into "0=MALE,1=FEMALE" or something similar.
For example, a node's metric might look like this at a given point in time:
2*ACT_SCORE + 0*GENDER + (-5)*VARISTY_SPORTS ...
The coefficients for each values are exactly what the network is searching for as it converges.
If you change the scale of a value, like ACT_SCORE, you just change the scale of the coefficient that will be found by the reciporical of that scaling factor.
The result should still be the same.
There are other concerns in terms of accuracy (computers have limited capacity to represent small fractions) and speed that may enter this, but not being familiar with NEURO XL, I can't say whether or not they apply for this technology.
Question 4
I am predicting the actual score, but in reality, I'm NOT that concerned about the exact score but more of a range. Would my network be more accurate if I grouped the output scores into buckets and then tried to predict this number instead of the actual score?
This will reduce accuracy, although you should converge to a solution much faster with fewer possible outputs (scores).
Neural Networks actually describe very high-dimensional functions in their input variables.
If you reduce the granularity of that function's output space, you essentially state that you don't care about local minima and maxima in that function, especially around the borders between your output scores.
As a result, you are sacrificing information that may be an essential component of the "true" function that you are searching for.
I hope this has been helpful, but you really should break this question down into its many components and ask them separately on different sites -- potentially some of them do belong here on StackOverflow as well.
I am trying to figure out how to generate a colored chart on a matlab, like the one you can find in here on page 9. (You will have to look through it to find what I am referring to - Stackoverflow doesn't allow me to post pictures in the postings just yet.)
A few questions:
I do have the table, but my table is a set of discreet points, not a continuous spectrum. So... can I do it in the first place?
If it IS possible, how would I do it?
(By the way, that table is from combat simulation for Risk - I am doing the combat simulation for Risk II, just for fun.)
The type of image you are looking for, as can be seen on Page 9, is a imagesc plot. Here's a simple example, using a double sin function. Done without vectorization for simplicity.
x=0:pi/180:pi;
y=0:pi/180:pi;
output=zeros(length(x),length(y));
for ix=1:length(x)
for iy=1:length(x)
output(ix,iy)=sin(x(ix)*2)*cos(y(iy)*4);
end
end
figure;imagesc(x,y,output)
I think you're looking for the filled contourplot.
See also: http://www.mathworks.nl/help/techdoc/ref/contourf.html
I have a picture.1200*1175 pixel.I want to train a net(mlp or hopfield) to learn a specific part of it(201*111pixel) to save its weight to use in a new net(with the same previous feature)only without train it to find that specific part.now there are this questions :what kind of nets is useful;mlp or hopfield,if mlp;the number of hidden layers;the trainlm function is unuseful because "out of memory" error.I convert the picture to a binary image,is it useful?
What exactly do you need the solution to do? Find an object with an image (like "Where's Waldo"?). Will the target object always be the same size and orientation? Might it look different because of lighting changes?
If you just need to find a fixed pattern of pixels within a larger image, I suggest using a straightforward correlation measure, such as crosscorrelation to find it efficiently.
If you need to contend with any of the issues mentioned above, then there are two basic solutions: 1. Build a model using examples of the object in different poses, scalings, etc. so that the model will recognize any of them, or 2. Develop a way to normalize the patch of pixels being examined, to minimize the effect of those distortions (like Hu's invariant moments). If nothing else, yuo'll want to perform some sort of data reduction to get the number of inputs down. Technically, you could also try a model which is invariant to rotations, etc., but I don't know how well those work. I suspect that they are more tempermental than traditional approaches.
I found AdaBoost to be helpful in picking out only important bits of an image. That, and resizing the image to something very tiny (like 40x30) using a Gaussian filter will speed it up and put weight on more of an area of the photo rather than on a tiny insignificant pixel.
I'm currently writing an optimization algorithm in MATLAB, at which I completely suck, therefore I could really use your help. I'm really struggling to find a good way of representing a graph (or well more like a tree with several roots) which would look more or less like this:
alt text http://img100.imageshack.us/img100/3232/graphe.png
Basically 11/12/13 are our roots (stage 0), 2x is stage1, 3x stage2 and 4x stage3. As you can see nodes from stageX are only connected to several nodes from stage(X+1) (so they don't have to be connected to all of them).
Important: each node has to hold several values (at least 3-4), one will be it's number and at least two other variables (which will be used to optimize the decisions).
I do have a simple representation using matrices but it's really hard to maintain, so I was wondering is there a good way to do it?
Second question: when I'm done with that representation I need to calculate how good each route (from roots to the end) is (like let's say I need to compare is 11-21-31-41 the best or is 11-21-31-42 better) to do that I will be using the variables that each node holds. But the values will have to be calculated recursively, let's say we start at 11 but to calcultate how good 11-21-31-41 is we first need to go to 41, do some calculations, go to 31, do some calculations, go to 21 do some calculations and then we can calculate 11 using all the previous calculations. Same with 11-21-31-42 (we start with 42 then 31->21->11). I need to check all the possible routes that way. And here's the question, how to do it? Maybe a BFS/DFS? But I'm not quite sure how to store all the results.
Those are some lengthy questions, but I hope I'm not asking you for doing my homework (as I got all the algorithms, it's just that I'm not really good at matlab and my teacher wouldn't let me to do it in java).
Granted, it may not be the most efficient solution, but if you have access to Matlab 2008+, you can define a node class to represent your graph.
The Matlab documentation has a nice example on linked lists, which you can use as a template.
Basically, a node would have a property 'linksTo', which points to the index of the node it links to, and a method to calculate the cost of each of the links (possibly with some additional property that describe each link). Then, all you need is a function that moves down each link, and brings the cost(s) with it when it moves back up.