Questions tagged [penalized]

Methods of modifying objective functions to control the solutions of optimization problems.

Filter by
Sorted by
Tagged with
0
votes
0answers
16 views

Interpretability of the lasso coefficients

I have implemented the lasso method to a data set with 1000+ variables. I have reported the MSE value on the test set and the number of nonzero variables. I need to report the interpretability of the ...
2
votes
0answers
100 views
+50

cross validation in lasso problem

Let us consider the following lasso estimator: $$ \hat{\beta}_{L} = \arg\min \, \frac{1}{n}\sum_{i}^{n}||y_{i} - \textbf{x}_{i}\beta||_{2}^{2} + \frac{\lambda_{n}}{n}\sum_{j=1}^{p}|\beta_{j}| $$ and ...
1
vote
1answer
47 views

Howe to perform ridge regression only on a subset of the variables

I am trying to code some algorithm that performs ridge-regression with penalty parameter $\lambda$ on all features except for a specific subset. Let $\mathbf{X}$ be the $n \times p$ matrix for $n$ ...
0
votes
0answers
4 views

Penalized least squares in R [migrated]

Disclaimer: It might be a purely R question, but i thought i'd post it here giving the goal. Using a smaller subset of movielens data , i'm trying to penalize sparsely rated movies by ...
1
vote
0answers
8 views

Is the SCAD a sort of BLUE estimator in high dimension?

Given a response vector $y$ and a data matrix $X$, the OLS estimator can be obtained by solving the minimization problem: $$ min_{\beta}\left\{\|y-X\beta \|_2^2\right\}$$ Under the OLS hypothesis, ...
3
votes
0answers
85 views

Bias correcting penalized maximum likelihood / maximum a posteriori estimates

Suppose an estimator $\hat\theta_T$ is defined as the value of $\theta$ maximizing: $$\sum_{t=1}^T{l(y_t|\theta)}+\mu_T g(\theta),$$ where $l(y_t|\theta)$ is the log-likelihood of observation $t$, $\...
3
votes
1answer
65 views

Does poisson penalized quasi likelihood regression use biased estimators?

A professor told me the Poisson glmmPQL (mixed-effects/hierarchical) regression gives biased estimates. The paper https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3886992/ from PLOS One says that ...
9
votes
1answer
727 views

Is there a difference between LASSO regularisation and LASSO penalisation?

I've seen the terms LASSO regularisation and LASSO penalisation used interchangeably? Is this correct, are they the same thing or what are the differences? Thanks!
5
votes
1answer
103 views

Smoothed Moments as Function of Predictor

Setup Let $x$ describe a continuous predictor variable (e.g. age). Let $Y$ be a random variable (e.g. height) which is some function of $x$. The data consists of $n$ points, each a combination of $x$...
0
votes
0answers
4 views

How to choose the two parameters of MCP penalty function?(a and λ)

As shown in the topic The MCP penalty function takes two arguments, $λ$ uses K folding for cross validation, and what about $a$?
0
votes
0answers
17 views

How many hash buckets should I set with penalized logistic regression over large dataset

I am working with a logistic regression, and one of the predictors I am considering is a truncated ip address--as a proxy for location. I have about 200 million rows in the dataset but there are about ...
1
vote
0answers
13 views

Random variables with known ratios

I have $n$ samples with $m$ random variables in each: $\{x_{i\alpha}|i=1..n, \alpha=1..m\}$. The variables are expected to obey known ratios (they are fractions of an unknown quantity), e.g. $x_{\...
1
vote
0answers
11 views

Non-differentiable regression penalties having non-zero probability of inducing sparsity

I was wondering if anyone has any analytic insight into why non-differentiable penalties can set coefficients to zero, and if there is a more relaxed alternative to this definition regarding the ...
1
vote
0answers
32 views

Lasso: more penalization with more data?

I am currently doing a backtest of a financial data set with an expanding window. For this, I estimate a Lasso model each month. Hence, each month that I estimate the model, I will have more data. ...
0
votes
0answers
28 views

How am I supposed to understand the penalty term to penalize lack of smoothness?

The penalty term is written as $\lambda\int f''(t)^2 dt$. I am told that if $f$ is wiggly, then $f''$ is big, so this term penalizes the lack of smoothness of $f$. Uh, okay, that makes sense ... but ...
1
vote
2answers
73 views

Regression penalisation to keep correlated variables

I am an avid user of logistic regression and more specifically of penalised logistic regression. With standard penalisation (Ridge, Lasso, Elasticnet) the goal is to avoid overfitting and translates ...
0
votes
0answers
24 views

robust regression in glmnet?

I have data with many errors in the input. However, I also have much more independent variables than samples. I have been using glmnet for penalized regression. Is there a good algorithm like glmnet ...
0
votes
0answers
127 views

Creating a risk score from Cox Regression

I have two datasets with palliative cancer patients including 106 and 60 patients, respectively. I have biomarkers of inflammation and coagulation, as well as clinical characteristics for all patients....
2
votes
1answer
72 views

Seeking to understand using the Firth correction in Generalized Estimating Equations to deal with quasi-complete separation

In order to deal with complete separation in my data someone suggested that I run penalized GEE (PGEE) by adding a Firth-type penalty term in R. Although I have read many papers on the Firth ...
0
votes
0answers
8 views

Appropriate machine learning technique for spectral data and low-frequency feedback

I have a performance measure and a data source that basically supplies a complex and varying waveform. I would like to apply some unstructured learning technique to try and find a pattern in the ...
0
votes
0answers
41 views

Elastic net regression with uneven penalties for predictors

For a regression model where you are certain that y that depends on some predictors but are agnostic about whether some other predictors should enter, how should you incorporate this prior information?...
1
vote
0answers
34 views

Is it useful to use sparse regression (e.g. Lasso) when the number of observations is significantly larger than the number of covariates?

I'm learning about penalized/sparse regression and I noticed that the examples used for penalized/sparse regression, e.g. Lasso, are usually cases where the number of observations is significantly ...
0
votes
0answers
13 views

In case of elastic-net regression, what is the updating equation of gamma0

In case of elastic-net regression, what is the updating equation of gamma0 I have a proble with dealing elastic-net regression for gamma and beta, ( argmin 1/2nSummation(yi-gamma-bxi)^2 + lambda(1-...
0
votes
0answers
21 views

Link between norm of weights/coefficients and smoothness

We often avoid overfitting by penalizing the norm of the weights/coefficients (in a classic Ridge or Lasso regression). I understand that we want smooth functions as they will be more likely to ...
1
vote
1answer
86 views

Why doesn't penalized cubic regression reduce the number of knots in a GAM?

As far as I understand, cubic regression penalization prevents overfitting by reducing the number of knots by penalizing wiggliness. The supplied parameter k serves ...
1
vote
0answers
99 views

LASSO regression: which method is better for selecting $\lambda$ in this case?

I am currently working on a method for adaptive knot placement in Spline regression. Following Osborne et.al. (1998), Yuan et.al. (2014) I am interested in using LASSO regression to select a subset of ...
0
votes
1answer
112 views

A reward becomes a penalty if

I am working to build a reinforcement agent with DQN. The agent would be able to place buy and sell orders for a day trading purpose. I am facing a little problem with that project. The question is "...
1
vote
1answer
108 views

why does lasso select at most n predictors?

From the seminal paper on elastic net regularization from Zou and Hastie 2005, I read ...
1
vote
0answers
69 views

Removing the intercept term for penalized logistic regression

I am working on lasso logistic regression and am trying to remove the intercept term from the penalty function. I have tried to use the mean centering theory but for logistic regression it can not be ...
2
votes
1answer
91 views

Large value of $X\beta$ in logistic regression?

In logistic regression, the probability is obtained from $$ Pr = \frac{\exp(X\beta)}{1 + \exp(X\beta)} ~~~~ (1) $$ From the plot below, it is obvious that if $X\beta$ > 10, the probability approaches ...
0
votes
1answer
71 views

Confidence limits for constrained penalized log likelihood model

I am estimating parameter $\beta$ as: \begin{align} \hat \beta &= \mathop{\mathrm{arg\,max}}_\beta \;\; l(\beta;X,y) - \frac{\lambda}{2}\left(\tilde y-g(\beta,\tilde X)\right)^\prime C^\prime C\...
0
votes
0answers
52 views

Adding additional constrains to OpenAi Gym

I'm currently working trough some examples which should finally end in a DQN Reinforcement Learning for the CartPole example in the openAI-Gym. Copied some code from GitHub which isn't deep yet: <...
3
votes
0answers
363 views

Calibration of penalized (LASSO or ELasticNet) logistic regression models

I would be very grateful for any help me with the following general query regarding calibration of penalized models with a binary outcome. I would like my prediction model to be calibrated (mean ...
2
votes
0answers
117 views

Elastic Net and collinearity

I am performing elastic net for variable selection on a dataset of 95 records and 41 variables. The response is a continuous numerical. I choose the alpha and lambda parameters through 10 fold cross ...
0
votes
1answer
310 views

Elastic net visualization [closed]

Is there a way to display in a graph the elastic net (or penalized regression in general) results? Specifically, how can I render the coefficients of the variables?
6
votes
1answer
747 views

Does Regularized Logistic Regression Produce Calibrated Results?

It has been asked and addressed here that logistic regression modelling is calibrated already and there is no need for calibration of it. To me it seems the argument provided there does not follow ...
1
vote
1answer
492 views

Sample size calculation for elastic net regression

I am using elastic net regression to investigate the effect of preditors on the response variable while accounting for multicollinearity among the predictors. But I wish to perform a sample size ...
3
votes
0answers
41 views

How to prove oracle properties in Fan and Li (2001) paper

I am studying Fan and Li's 2001 paper "Variable selection via nonconcave penalized Likelihood andits oracle properties" but I am having troubles understanding Theorem 1 proof (page 1359). I follow the ...
3
votes
1answer
529 views

Interpreting glmnet Lasso coefficients on dummy variables (multiple levels) [duplicate]

I am trying to apply glmnet's lasso to a set of features in which there are multiple categorical variables with multiple levels. My intention is to let lasso reduce ...
4
votes
0answers
77 views

Using covariates from penalized regression model in unpenalized model

The good news where I am is that researchers are doing less stepwise covariate selection now that I've introduced penalized regression. The bad news is that researchers want to use elastic-net ...
1
vote
0answers
181 views

In LASSO, does it make sense to choose lambda based on the mean error associated with different lambda values, over multiple cross-validations?

I am running a LASSO regression, but am put off by the different values of lambda each time I run the cross-validation. Does it make sense to run a cross validation multiple times, take the mean error ...
12
votes
1answer
2k views

Why is R Squared not a good measure for regressions fit using LASSO?

I have read in several places that R Squared is not an ideal measure when a model is fit using LASSO. However, I'm not clear on exactly why that is. In addition, could you recommend the best ...
0
votes
0answers
122 views

Equivalent of using a Poisson prior in terms of a penalized regression?

I know that most penalized regressions have also a Bayesian interpretation, e.g. ridge least squares regression corresponds to the MAP estimate obtained under a Gaussian prior in a Bayesian regression,...
0
votes
0answers
136 views

How do you explain many optimal models in penalized Logistic regression?

I am building penalized logistic regression using Lasso and Ridge methods. I know that the best model chosen by the program is which has alpha = 1 and lambda = 0.06....
1
vote
0answers
103 views

“beta_given” column in the h2o.glm beta_constraints

What does the "beta_given" column do in the h2o.glm beta_constraints parameter? h2o is an open source library for machine learning algorithms. There are several online examples on how to install the ...
1
vote
0answers
39 views

Comparing shrunk model vs null model?

I have a null model $M_0$ with parameter $\theta$ and a more general model $M_1$ with parameters $(\theta,\alpha)$. I know that $M_0=M_1$ for $\alpha=0$. I am estimating $M_1$ using a penalized ...
6
votes
1answer
505 views

How is the minimum $\lambda$ computed in group LASSO?

The LASSO problem works by minimizing $$\min_\beta (\frac{1}{2}\left\rVert y-X\beta\right\rVert^2_2+\lambda\left\rVert\beta\right\rVert_1)$$ Here in this webpage I found that the minimal value of ...
1
vote
0answers
17 views

Elastic net with increased penalty for lower quality features

I’m building multinomial classification models using features characterized with high false-positive rate. Meaning, as the signal rate of the feature is lower (say gene expression abundance) the more ...
4
votes
1answer
149 views

How to find smallest $\lambda$ such all lasso coefficients are set to 0, depending on the intercept?

While bouilding a LASSO-penalized model it is well known that $\lambda =\left\lVert X^ty\right\lVert_\infty$ is the minimum value for which all the $\beta$ coefficients of the model are 0. Consider ...
1
vote
0answers
13 views

Bayesian Information Criteria variable penalisation for increasing N

I am trying to understand the Bayesian Information Criteria for model performance, given by the formula below: BIC=−2ln(L)+kln(n) I understand the aim is to minimize this function. I can see that ...