Timezone: »

 
Statistical and Computational Complexities of BFGS Quasi-Newton Method for Generalized Linear Models
Qiujiang Jin · Aryan Mokhtari · Nhat Ho · Tongzheng Ren

@
in Order up! The Benefits of Higher-Order Optimization in Machine Learning »

The gradient descent (GD) method has been used widely to solve parameter estimation in generalized linear models (GLMs), a generalization of linear models when the link function can be non-linear. While GD has optimal statistical and computational complexities for estimating the true parameter under the high signal-to-noise ratio (SNR) regime of the GLMs, it has sub-optimal complexities when the SNR is low, namely, the iterates of GD require polynomial number of iterations to reach the final statistical radius. The slow convergence of GD for the low SNR case is mainly due to the local convexity of the least-square loss functions of the GLMs. To address the shortcomings of GD, we propose to use the BFGS quasi-Newton method to solve parameter estimation of the GLMs. On the optimization side, when the SNR is low, we demonstrate that iterates of BFGS converge linearly to the optimal solution of the population least-square loss function. On the statistical side, we prove that the iterates of BFGS reach the final statistical radius of the low SNR GLMs after a logarithmic number of iterations, which is much lower than the polynomial number of iterations of GD. We also present numerical experiments that match our theoretical findings.

[ Poster] 

Author Information

Qiujiang Jin (University of Texas, Austin)
Aryan Mokhtari (UT Austin)
Nhat Ho (University of Texas at Austin)
Tongzheng Ren (UT Austin)

More from the Same Authors