I am building in Python a credit scorecard using this public dataset: https://www.kaggle.com/sivakrishna3311/delinquency-telecom-dataset
It's a binary classification problem:
Target = 1 -> Good applicant
Target = 0 -> Bad applicant
I only have numeric continuous predictive characteristics.
In the credit industry it is a legal requirement to explain why an applicant got rejected (or didn't even get the maximum score): to meet that requirement, Adverse Codes are produced.
In a classic logistic regression approach, one would do this:
calculate the Weight-of-Evidence (WoE) for each predictive
characteristic (forcing a monotonic relationship between the feature
values and the WoE or log(odds)). In the following example, the
higher the network Age the higher the Weight-of-Evidence (WoE):
replace the data values with the correspondent Weight-of-Evidence.
For example, a value of 250 for Network Age would be replaced by
0.04 (which is the correspondent WoE).
Train a logistic regression
After some linear transformations you'd get something like this:
And therefore it'd be straightforward to assign the Adverse Codes, so that the bin with the maximum score doesn't return an Adverse Code. For example:
Now, I want to train an XGBoost (which typically outperforms a logistic regression on a imbalanced, low noise data). XGBoost are very predictive but need to be explained (typically via SHAP).
What I have read is that in order to make the model decision explainable you must ensure that the monotonic constraints are applied.
Question 1. Does it mean that I need to train the XGBoost on the Weight-of-Evidence transformed data like it's done with the Logistic Regression (see point 2 above)?
Question 2. In Python, the XGBoost package offers the option to set monotonic constraints (via the monotone_constraints option). If I don't transform the data by replacing the Weight-of-Evidence (therefore removing all monotonic constraints) does it still make sense to use "monotone_constraints" in XGboost for a binary problem? I mean, does it make sense to use monotone_constraints with a XGBClassifier at all?
Thanks.
I am trying to develop Cox PH model with time-varying covariates in R. I use coxph function from survival package. There was not any trouble during estimation process, though coefficient value of one covariates is too large, in particular, 2.5e+32.
I can't guess what is reason of this problem and how to tackle it. This variable is nonstationary and proportional assumption is violated. Does either of this facts may cause such a big value of coefficient?
More information could help framing your problem.
Anyway, I doubt non-proportionality is to blame. It would imply that you have some outliers heavily biasing your coefficient beyond reasonable expectations. You could give this a quick look by plotting the output of cox.zph.
Another possible explanation is that this rather depends on the unit of measure you used to define your covariate. Can the magnitude of the coefficient be meaningfully interpreted? If so, you could simply re-scale/standardise/log-transform that covariate to obtain a 'more manageable' coefficient (if this is theoretically appropriate).
This could also be due to the so called 'complete separation', which has been discussed here and here.
I am using the R package gbm to fit probabilistic classifiers in a dataset with > 2 classes. I am using distribution = "multinomial" as argument, however, I have some difficulties to find out implementation details of what is actually implemented by that choice.
The help function for gbm states that
Currently available options are "gaussian" (squared error), "laplace" (absolute loss), "tdist" (t-distribution loss), "bernoulli" (logistic regression for 0-1 outcomes), "huberized" (huberized hinge loss for 0-1 outcomes), classes), "adaboost" (the AdaBoost exponential loss for 0-1 outcomes), "poisson" (count outcomes), "coxph" (right censored observations), "quantile", or "pairwise" (ranking measure using the LambdaMart algorithm).
and does not list multinomial, whereas the paragraph preceding the one I copied states that
... If not specified, gbm will try to guess: ... if the response is a factor,multinomial is assumed; ...
I would like to know which loss function is implemented if I specify distribution = "multinomial". The documentation in the vignette which can be accessed via
utils::browseVignettes("gbm")
does not contain the word "multinomial" or any descriptions of what that argument implies.
I have tried to look at the package source code, but can't find the information there as well. It seems that the relevant things happen in the C++ functions in the file /src/multinomial.cpp, however, my knowledge of C++ is too limited to understand what is going on there.
Background
The purpose of a generalized linear model, as I understand it, is to allow one to fit coefficients of a linear model in the presence of noise that is not normally distributed.
The 'glmulti' package allows one to evaluate many candidate coefficient combinations to find the one that has highest yield in fit (minimum error) for lowest cost (number of coefficients) using information criteria like akaike information criterion (aic) or Bayes information criterion (bic).
For my case, I suspect that instead of being a purely linear expression of the variables, I may have a rational expression.
Question
Is there a clean way to fit this using an off-the-shelf package for R? I know that the MatLab (or MATLAB) curve fit toolbox has a rational fit, but that requires two very large purchases. If not in R, is there an option using another language like python, sage, or similar?
In the output layer of a neural network, it is typical to use the softmax function to approximate a probability distribution:
This is expensive to compute because of the exponents. Why not simply perform a Z transform so that all outputs are positive, and then normalise just by dividing all outputs by the sum of all outputs?
There is one nice attribute of Softmax as compared with standard normalisation.
It react to low stimulation (think blurry image) of your neural net with rather uniform distribution and to high stimulation (ie. large numbers, think crisp image) with probabilities close to 0 and 1.
While standard normalisation does not care as long as the proportion are the same.
Have a look what happens when soft max has 10 times larger input, ie your neural net got a crisp image and a lot of neurones got activated
>>> softmax([1,2]) # blurry image of a ferret
[0.26894142, 0.73105858]) # it is a cat perhaps !?
>>> softmax([10,20]) # crisp image of a cat
[0.0000453978687, 0.999954602]) # it is definitely a CAT !
And then compare it with standard normalisation
>>> std_norm([1,2]) # blurry image of a ferret
[0.3333333333333333, 0.6666666666666666] # it is a cat perhaps !?
>>> std_norm([10,20]) # crisp image of a cat
[0.3333333333333333, 0.6666666666666666] # it is a cat perhaps !?
I've had this question for months. It seems like we just cleverly guessed the softmax as an output function and then interpret the input to the softmax as log-probabilities. As you said, why not simply normalize all outputs by dividing by their sum? I found the answer in the Deep Learning book by Goodfellow, Bengio and Courville (2016) in section 6.2.2.
Let's say our last hidden layer gives us z as an activation. Then the softmax is defined as
Very Short Explanation
The exp in the softmax function roughly cancels out the log in the cross-entropy loss causing the loss to be roughly linear in z_i. This leads to a roughly constant gradient, when the model is wrong, allowing it to correct itself quickly. Thus, a wrong saturated softmax does not cause a vanishing gradient.
Short Explanation
The most popular method to train a neural network is Maximum Likelihood Estimation. We estimate the parameters theta in a way that maximizes the likelihood of the training data (of size m). Because the likelihood of the whole training dataset is a product of the likelihoods of each sample, it is easier to maximize the log-likelihood of the dataset and thus the sum of the log-likelihood of each sample indexed by k:
Now, we only focus on the softmax here with z already given, so we can replace
with i being the correct class of the kth sample. Now, we see that when we take the logarithm of the softmax, to calculate the sample's log-likelihood, we get:
, which for large differences in z roughly approximates to
First, we see the linear component z_i here. Secondly, we can examine the behavior of max(z) for two cases:
If the model is correct, then max(z) will be z_i. Thus, the log-likelihood asymptotes zero (i.e. a likelihood of 1) with a growing difference between z_i and the other entries in z.
If the model is incorrect, then max(z) will be some other z_j > z_i. So, the addition of z_i does not fully cancel out -z_j and the log-likelihood is roughly (z_i - z_j). This clearly tells the model what to do to increase the log-likelihood: increase z_i and decrease z_j.
We see that the overall log-likelihood will be dominated by samples, where the model is incorrect. Also, even if the model is really incorrect, which leads to a saturated softmax, the loss function does not saturate. It is approximately linear in z_j, meaning that we have a roughly constant gradient. This allows the model to correct itself quickly. Note that this is not the case for the Mean Squared Error for example.
Long Explanation
If the softmax still seems like an arbitrary choice to you, you can take a look at the justification for using the sigmoid in logistic regression:
Why sigmoid function instead of anything else?
The softmax is the generalization of the sigmoid for multi-class problems justified analogously.
I have found the explanation here to be very good: CS231n: Convolutional Neural Networks for Visual Recognition.
On the surface the softmax algorithm seems to be a simple non-linear (we are spreading the data with exponential) normalization. However, there is more than that.
Specifically there are a couple different views (same link as above):
Information Theory - from the perspective of information theory the softmax function can be seen as trying to minimize the cross-entropy between the predictions and the truth.
Probabilistic View - from this perspective we are in fact looking at the log-probabilities, thus when we perform exponentiation we end up with the raw probabilities. In this case the softmax equation find the MLE (Maximum Likelihood Estimate)
In summary, even though the softmax equation seems like it could be arbitrary it is NOT. It is actually a rather principled way of normalizing the classifications to minimize cross-entropy/negative likelihood between predictions and the truth.
The values of q_i are unbounded scores, sometimes interpreted as log-likelihoods. Under this interpretation, in order to recover the raw probability values, you must exponentiate them.
One reason that statistical algorithms often use log-likelihood loss functions is that they are more numerically stable: a product of probabilities may be represented be a very small floating point number. Using a log-likelihood loss function, a product of probabilities becomes a sum.
Another reason is that log-likelihoods occur naturally when deriving estimators for random variables that are assumed to be drawn from multivariate Gaussian distributions. See for example the Maximum Likelihood (ML) estimator and the way it is connected to least squares.
We are looking at a multiclass classification problem. That is, the predicted variable y can take one of k categories, where k > 2. In probability theory, this is usually modelled by a multinomial distribution. Multinomial distribution is a member of exponential family distributions. We can reconstruct the probability P(k=?|x) using properties of exponential family distributions, it coincides with the softmax formula.
If you believe the problem can be modelled by another distribution, other than multinomial, then you could reach a conclusion that is different from softmax.
For further information and a formal derivation please refer to CS229 lecture notes (9.3 Softmax Regression).
Additionally, a useful trick usually performs to softmax is: softmax(x) = softmax(x+c), softmax is invariant to constant offsets in the input.
The choice of the softmax function seems somehow arbitrary as there are many other possible normalizing functions. It is thus unclear why the log-softmax loss would perform better than other loss alternatives.
From "An Exploration of Softmax Alternatives Belonging to the Spherical Loss Family" https://arxiv.org/abs/1511.05042
The authors explored some other functions among which are Taylor expansion of exp and so called spherical softmax and found out that sometimes they might perform better than usual softmax.
I think one of the reasons can be to deal with the negative numbers and division by zero, since exp(x) will always be positive and greater than zero.
For example for a = [-2, -1, 1, 2] the sum will be 0, we can use softmax to avoid division by zero.
Adding to Piotr Czapla answer, the greater the input values, the greater the probability for the maximum input, for same proportion and compared to the other inputs:
Suppose we change the softmax function so the output activations are given by
where c is a positive constant. Note that c=1 corresponds to the standard softmax function. But if we use a different value of c we get a different function, which is nonetheless qualitatively rather similar to the softmax. In particular, show that the output activations form a probability distribution, just as for the usual softmax. Suppose we allow c to become large, i.e., cāā. What is the limiting value for the output activations a^L_j? After solving this problem it should be clear to you why we think of the c=1 function as a "softened" version of the maximum function. This is the origin of the term "softmax". You can follow the details from this source (equation 83).
While it indeed somewhat arbitrary, the softmax has desirable properties such as:
being easily diferentiable (df/dx = f*(1-f))
when used as the output layer for a classification task, the in-fed scores are interpretable as log-odds