Any binary one-hot encoding is aware of only values seen in training, so features not encountered during fitting will be silently ignored. For real time, where you have millions of records in a second, and features have very high cardinality, you need to keep your hasher/mapper updated with the data.
How can we do an incremental update to the hasher (rather calculating the entire fit() every time we incounter a new feature-value pair)? What is the suggested approach here the tackle this?
It depends on the learning algorithm that you are using. If you are using a method that has been designated for sparse data sets (FTRL, FFM, linear SVM) one possible approach is the following (note that it will introduce collisions in the features and a lot of constant columns).
First allocate for each element of your sample a (as large as possible) vector V, of length D.
For each categorical variable, evaluate hash(var_name + "_" + var_value) % D. This gives you an integer i, and you can store V[i] = 1.
Therefore, V never grows larger as new features appear. However, as soon as the number of features is large enough, some features will collide (i.e. be written at the same place) and this may result in an increased error rate...
Edit. You can write your own vectorizer to avoid collisions. First call L the current number of features. Prepare the same vector V of length 2L (this 2 will allow you to avoid collisions as new features arrive - at least for some time, depending of the arrival rate of new features).
Starting with an emty dictionary<input_type,int>, associate to each feature an integer. If have already seen the feature, return the int corresponding to the feature. If not, create a new entry with an integer corresponding to the new index. I think (but I am not sure) this is what LabelEncoder does for you.
Related
I have many sequences of data looking like this:
s1 = t11, t12, ..., t1m_1
s2 = t21, t22, ..., t2m_2
...
si = ti1, ti2, ..., tim_i
si means the i-th sequence, tij means the i-th sequence be accessed at time tj
each sequence has different length of data (m_1 may not equal to m_2),
and each sequence's data means that the sequence si was accessed time at ti1, ti2, ..., tim_i.
My goal is to cluster the similar access-time sequences.
I'm not sure whether I can translate this problem to a time-series problem.
For my understanding the time-series data like that each sequence's data means the value at that time like stock data, but my sequence's value means which time the sequence be accessed.
If it can translate to time-series problem, but there is another problem. The problem is that the sequence's access time is very discrete (may be accessed at 1s, 1000s, 2000s), so if I translate to time-series format, its space would be very large, I think this can't run cluster with some algorithm like (DTW), its time complexity may too large.
As you pointed out, DTW would be quite slow, since comparing the first two series takes k * m_1 * m_2 operations.
To avoid this, and to more easily compare your sequences, you might somehow hammer them into the same format (thereby also losing information).
Here are some ideas:
Differentiate to obtain times-between-accesses, and build histograms with fixed bins across all data.
Count the number of accesses during each minute every week (and divide by number of times that minute-of-week appears in each series). Adapt to timescales of interest.
Count "number of accesses up until now". So, instead of having data points only when an access was made ("sparse"), you'd get a data point for every timestamp ("dense") showing accesses for every minute up to the current one.
#3 would be similar to an "integral image" in computer vision. After this, new summarization techniques open up, like moving averages, or even direct comparison (if the recordings happen in parallel).
In order to pick a more useful representation, you need to think about what is meaningful in your application.
After you get a uniform-length representation, you can use cheaper similarity measures. A typical one is cosine similarity (but be sure to normalize first).
Data: When I have N rows of data like this: (x,y,z) where logically f(x,y)=z, that is z is dependent on x and y, like in my case (setting1, setting2 ,signal) . Different x's and y's can lead to the same z, but the z's wouldn't mean the same thing.
There are 30 unique setting1, 30 setting2 and 1 signal for each (setting1, setting2)-pairing, hence 900 signal values.
Data set: These [900,3] data points are considered 1 data set. I have many samples of these data sets.
I want to make a classification based on these data sets, but I need to flatten the data (make them all into one row). If I flatten it, I will duplicate all the setting values (setting1 and setting2) 30 times, i.e. I will have a row with 3x900 columns.
Question:
Is it correct to keep all the duplicate setting1,setting2 values in the data set? Or should I remove them and only include the unique values a single time?, i.e. have a row with 30 + 30 + 900 columns. I'm worried, that the logical dependency of the signal to the settings will be lost this way. Is this relevant? Or shouldn't I bother including the settings at all (e.g. due to correlations)?
If I understand correctly, you are training NN on a sample where each observation is [900,3].
You are flatning it and getting an input layer of 3*900.
Some of those values are a result of a function on others.
It is important which function, as if it is a liniar function, NN might not work:
From here:
"If inputs are linearly dependent then you are in effect introducing
the same variable as multiple inputs. By doing so you've introduced a
new problem for the network, finding the dependency so that the
duplicated inputs are treated as a single input and a single new
dimension in the data. For some dependencies, finding appropriate
weights for the duplicate inputs is not possible."
Also, if you add dependent variables you risk the NN being biased towards said variables.
E.g. If you are running LMS on [x1,x2,x3,average(x1,x2)] to predict y, you basically assign a higher weight to the x1 and x2 variables.
Unless you have a reason to believe that those weights should be higher, don't include their function.
I was not able to find any link to support, but my intuition is that you might want to decrease your input layer in addition to omitting the dependent values:
From professor A. Ng's ML Course I remember that the input should be the minimum amount of values that are 'reasonable' to make the prediction.
Reasonable is vague, but I understand it so: If you try to predict the price of a house include footage, area quality, distance from major hub, do not include average sun spot activity during the open home day even though you got that data.
I would remove the duplicates, I would also look for any other data that can be omitted, maybe run PCA over the full set of Nx[3,900].
Any idea on the recommended parameters for OpenCV RTrees? I have read the documentation and I'm trying to apply it to MNIST dataset, i.e. 60000 training images, with 10000 testing images. I'm trying to optimize MaxDepth, MinSampleCount, setMaxCategories, and setPriors? e.g.
Ptr<RTrees> model = RTrees::create();
/* Depth of the tree.
A low value will likely underfit and conversely
a high value will likely overfit.
The optimal value can be obtained using cross validation
or other suitable methods.
*/
model->setMaxDepth(?); // letter_recog.cpp uses 10
/* minimum samples required at a leaf node for it to be split.
A reasonable value is a small percentage of the total data e.g. 1%.
MNIST 70000 * 0.01 = 700
*/
model->setMinSampleCount(700?); letter_recog.cpp uses 10
/* regression_accuracy – Termination criteria for regression trees.
If all absolute differences between an estimated value in a node and
values of train samples in this node are less than this parameter
then the node will not be split. */
model->setRegressionAccuracy(0); // I think this is already correct
/*
use_surrogates – If true then surrogate splits will be built.
These splits allow to work with missing data and compute variable importance correctly.'
To compute variable importance correctly, the surrogate splits must be enabled in
the training parameters, even if there is no missing data.
*/
model->setUseSurrogates(true); // I think this is already correct
/*
Cluster possible values of a categorical variable into K \leq max_categories clusters
to find a suboptimal split. If a discrete variable, on which the training procedure
tries to make a split, takes more than max_categories values, the precise best subset
estimation may take a very long time because the algorithm is exponential.
Instead, many decision trees engines (including ML) try to find sub-optimal split
in this case by clustering all the samples into max_categories clusters that is
some categories are merged together. The clustering is applied only in n>2-class
classification problems for categorical variables with N > max_categories possible values.
In case of regression and 2-class classification the optimal split can be found
efficiently without employing clustering, thus the parameter is not used in these cases.
*/
model->setMaxCategories(?); letter_recog.cpp uses 15
/*
priors – The array of a priori class probabilities, sorted by the class label value.
The parameter can be used to tune the decision tree preferences toward a certain class.
For example, if you want to detect some rare anomaly occurrence, the training base will
likely contain much more normal cases than anomalies, so a very good classification
performance will be achieved just by considering every case as normal.
To avoid this, the priors can be specified, where the anomaly probability is
artificially increased (up to 0.5 or even greater), so the weight of the misclassified
anomalies becomes much bigger, and the tree is adjusted properly. You can also think about
this parameter as weights of prediction categories which determine relative weights that
you give to misclassification. That is, if the weight of the first category is 1 and
the weight of the second category is 10, then each mistake in predicting the
second category is equivalent to making 10 mistakes in predicting the first category.
*/
model->setPriors(Mat()); // ?
/* If true then variable importance will be calculated and
then it can be retrieved by CvRTrees::get_var_importance().
*/
model->setCalculateVarImportance(true); // I think this is already correct
/*
The size of the randomly selected subset of features at each tree node and
that are used to find the best split(s). If you set it to 0 then the size
will be set to the square root of the total number of features.
*/
model->setActiveVarCount(0); // I think this is already correct
/*
CV_TERMCRIT_ITER Terminate learning by the max_num_of_trees_in_the_forest;
CV_TERMCRIT_EPS Terminate learning by the forest_accuracy;
CV_TERMCRIT_ITER | CV_TERMCRIT_EPS Use both termination criteria.
*/
model->setTermCriteria(TC(100,0.01f)); // I think this is already correct
I have a large dataset I am trying to do cluster analysis on using SOM. The dataset is HUGE (~ billions of records) and I am not sure what should be the number of neurons and the SOM grid size to start with. Any pointers to some material that talks about estimating the number of neurons and grid size would be greatly appreciated.
Thanks!
Quoting from the som_make function documentation of the som toolbox
It uses a heuristic formula of 'munits = 5*dlen^0.54321'. The
'mapsize' argument influences the final number of map units: a 'big'
map has x4 the default number of map units and a 'small' map has
x0.25 the default number of map units.
dlen is the number of records in your dataset
You can also read about the classic WEBSOM which addresses the issue of large datasets
http://www.cs.indiana.edu/~bmarkine/oral/self-organization-of-a.pdf
http://websom.hut.fi/websom/doc/ps/Lagus04Infosci.pdf
Keep in mind that the map size is also a parameter which is also application specific. Namely it depends on what you want to do with the generated clusters. Large maps produce a large number of small but "compact" clusters (records assigned to each cluster are quite similar). Small maps produce less but more generilized clusters. A "right number of clusters" doesn't exists, especially in real world datasets. It all depends on the detail which you want to examine your dataset.
I have written a function that, with the data set as input, returns the grid size. I rewrote it from the som_topol_struct() function of Matlab's Self Organizing Maps Toolbox into a R function.
topology=function(data)
{
#Determina, para lattice hexagonal, el número de neuronas (munits) y su disposición (msize)
D=data
# munits: número de hexágonos
# dlen: número de sujetos
dlen=dim(data)[1]
dim=dim(data)[2]
munits=ceiling(5*dlen^0.5) # Formula Heurística matlab
#munits=100
#size=c(round(sqrt(munits)),round(munits/(round(sqrt(munits)))))
A=matrix(Inf,nrow=dim,ncol=dim)
for (i in 1:dim)
{
D[,i]=D[,i]-mean(D[is.finite(D[,i]),i])
}
for (i in 1:dim){
for (j in i:dim){
c=D[,i]*D[,j]
c=c[is.finite(c)];
A[i,j]=sum(c)/length(c)
A[j,i]=A[i,j]
}
}
VS=eigen(A)
eigval=sort(VS$values)
if (eigval[length(eigval)]==0 | eigval[length(eigval)-1]*munits<eigval[length(eigval)]){
ratio=1
}else{
ratio=sqrt(eigval[length(eigval)]/eigval[length(eigval)-1])}
size1=min(munits,round(sqrt(munits/ratio*sqrt(0.75))))
size2=round(munits/size1)
return(list(munits=munits,msize=sort(c(size1,size2),decreasing=TRUE)))
}
hope it helps...
Iván Vallés-Pérez
I don't have a reference for it, but I would suggest starting off by using approximately 10 SOM neurons per expected class in your dataset. For example, if you think your dataset consists of 8 separate components, go for a map with 9x9 neurons. This is completely just a ballpark heuristic though.
If you'd like the data to drive the topology of your SOM a bit more directly, try one of the SOM variants that change topology during training:
Growing SOM
Growing Neural Gas
Unfortunately these algorithms involve even more parameter tuning than plain SOM, but they might work for your application.
Kohenon has written on the issue of selecting parameters and map size for SOM in his book "MATLAB Implementations and Applications of the Self-Organizing Map". In some cases, he suggest the initial values can be arrived at after testing several sizes of the SOM to check that the cluster structures were shown with sufficient resolution and statistical accuracy.
my suggestion would be the following
SOM is distantly related to correspondence analysis. In statistics, they use 5*r^2 as a rule of thumb, where r is the number of rows/columns in a square setup
usually, one should use some criterion that is based on the data itself, meaning that you need some criterion for estimating the homogeneity. If a certain threshold would be violated, you would need more nodes. For checking the homogeneity you would need some records per node. Agai, from statistics you could learn that for simple tests (small number of variables) you would need around 20 records, for more advanced tests on some variables at least 8 records.
remember that the SOM represents a predictive model. So validation is the key, absolutely mandatory. Yet, validation of predictive models (see typeI / II error entry in Wiki) is a subject on its own. And the acceptable risk as well as the risk structure also depend fully on your purpose.
You may test the dynamics of the error rate of the model by reducing its size more and more. Then take the smallest one with acceptable error.
It is a strength of the SOM to allow for empty nodes. Yet, there should not be too much of them. Let me say, less than 5%.
Taken all together, from experience, I would recommend the following criterion a minimum of the absolute number of 8..10 records, but those should not be more than 5% of all clusters.
Those 5% rule is of of course a heuristics, which however can be justified by the general usage of the confidence level in statistical tests. You may choose any percentage from 1% to 5%.
How can algorithms which partition a space in to halves, such as Suport Vector Machines, be generalised to label data with labels from sets such as the integers?
For example, a support vector machine operates by constructing a hyperplane and then things 'above' the hyperplane take one label, and things below it take the other label.
How does this get generalised so that the labels are, for example, integers, or some other arbitrarily large set?
One option is the 'one-vs-all' approach, in which you create one classifier for each set you want to partition into, and select the set with the highest probability.
For example, say you want to classify objects with a label from {1,2,3}. Then you can create three binary classifiers:
C1 = 1 or (not 1)
C2 = 2 or (not 2)
C3 = 3 or (not 3)
If you run these classifiers on a new piece of data X, then they might return:
C1(X) = 31.6% chance of being in 1
C2(X) = 63.3% chance of being in 2
C3(X) = 89.3% chance of being in 3
Based on these outputs, you could classify X as most likely being from class 3. (The probabilities don't add up to 1 - that's because the classifiers don't know about each other).
If your output labels are ordered (with some kind of meaningful, rather than arbitrary ordering). For example, in finance you want to classify stocks into {BUY, SELL, HOLD}. Although you can't legitimately perform a regression on these (the data is ordinal rather than ratio data) you can assign the values of -1, 0 and 1 to SELL, HOLD and BUY and then pretend that you have ratio data. Sometimes this can give good results even though it's not theoretically justified.
Another approach is the Cramer-Singer method ("On the algorithmic implementation of multiclass kernel-based vector machines").
Svmlight implements it here: http://svmlight.joachims.org/svm_multiclass.html.
Classification into an infinite set (such as the set of integers) is called ordinal regression. Usually this is done by mapping a range of continuous values onto an element of the set. (see http://mlg.eng.cam.ac.uk/zoubin/papers/chu05a.pdf, Figure 1a)