Softwares

Sat, Mar 14, 2020

R packages

1. The Python codes “REGS” is proposed for sampling from unnormalized densties by the techniques including Wasserstein gradient flows, numerical ODEs, density-ratio estimation and deep neural networks. REGS achieves a fantastic numerical performance on 2D mixtrues of Gaussian distributions with a large number of modes, a small variance, and a large distance between any two modes. “REGS” is available at https://github.com/xliusufe/REGS.

   • Refer to Feng X., Gao, Y., Huang, J., Jiao, Y. and Liu, X.* (2021). Relative Entropy Gradient Sampler for Unnormalized Distributions. Submitted.

2. The R-package “rbs” select the response and estimate regression coefficients simultaneously for multivariate regression with high-dimensional response variables. “rbs” is available at https://github.com/xliusufe/rbs.

   • Refer to Hu, J., Huang, J., Liu, X. and Liu, X.* (2022). Response Best-subset Selector for Multivariate Regression with High-dimensional Response Variables. Biometrika. Accepted.

3. The R-package “pqr” constructs confidence intervals of the coefficients of high-dimensional quantile regression via the regularized projection score estimation for the treatment effects. “pqr” is available at https://github.com/xliusufe/pqr.

   • Refer to Feng, X., Huang, J. and Liu, X.* (2020). Regularized projection score estimation of treatment effects in high-dimensional quantile regression. Statistica Sinica.

4. The R-package “tensorMQR” estimates the coefficients as a symmetric tensor for high-dimensional multiresponse quadratic regression models incorporating Tucker decomposition for the symmetric tensor and the steepest gradient descent algorithm on Stiefel manifold. “tensorMQR” is available at https://github.com/xliusufe/tensorMQR.

   • Refer to Liu, et al. (2020). “Symmetric tensor estimation for quadratic regression”.

5. The R-package “tensorMam” estimates the nonparametric curve for high-dimensional multivariate additive models incorporating Tucker decomposition for the tensor consisting of the coefficients. “tensorMam” is available at https://github.com/xliusufe/tensorMam.

   • Refer to Liu, X., Lian, H. and Huang, J. (2020). “A tensor estimation approach to multivariate additive models”.

6. The R-package “RidgeVar” estimates the err variance for high-dimensional linear regression in weak signal case. “RidgeVar” is available at https://github.com/xliusufe/RidgeVar.

   • Refer to Liu, X., Zheng, S. and Feng, X.* (2020). “Estimation of error variance via ridge regression”. Biometrika.

7. The R-package “IVGC” estimates coefficients for high-dimensional linear regression with instrument variable incorporating network structure. “IVGC” is available at https://github.com/xliusufe/IVGC.

   • Refer to Gao, et al. (2019). “Integrative analysis of genetical genomics data incorporating network structures”. Biometrics.

8. The R-package “plvs” estimates coefficients for high-dimensional quantile regression, including composite quantile, which is implemented by combining coordinate descent and MM algorithm. “plvs” is available at https://github.com/xliusufe/plvs.

   • Refer to Liu, et al. (2018). “Ultra-high dimensional variable selection for piecewise linear loss functions”. Manuscript

9. The R-package “FactSum” calculates the factorial of a large positive integer, that is n!, which may be much greater than the maximum memory of any data type defined by C/C++ or R. FactSum implements dramatically fast. It takes only 0.45 seconds to compute 10000! (it approximates 2.8E+35660), and 0.98 seconds to compute 10000! and sum=1!+2!+3!+…+10000! simultaneously. “FactSum” is available at https://github.com/xliusufe/FactSum. A web-based calculator can be found HERE. It is developed for teaching “Computer Programming”.

10. The R-package “sqrtn” calculates $\sqrt{n}$ with very high precision, where n is a positive integer. “sqrtn” implements dramatically fast. It takes only less than 30 seconds to approximate $\sqrt{2}$ with 100,000 digits. “sqrtn” is available at https://github.com/xliusufe/sqrtn. A web-based calculator can be found HERE. It is developed for teaching “Computer Programming”.

11. The R-package “PI” approximates $\pi$ with very high-precesion. It takes only 0.04 seconds to approximate $\pi$ with 1,000,000,000 digits. “PI” is available at https://github.com/xliusufe/PI. A web-based calculator can be found HERE. It is developed for teaching “Computer Programming”.

---

Homepage

• go back to
• go back to