asdfghjkl 0.1a2

ASDL: Automatic Second-order Differentiation (for Fisher, Gradient covariance, Hessian, Jacobian, and Kernel) Library

ASD(FGHJK)L

The library is called "ASDL", which stands for Automatic Second-order Differentiation (for Fisher, Gradient covariance, Hessian, Jacobian, and Kernel) Library. ASDL is a PyTorch extension for computing 1st/2nd-order metrics and performing 2nd-order optimization of deep neural networks.

You can import asdfghjkl by sliding your finger on a QWERTY keyboard :innocent:

import asdfghjkl

ADL vs ASDL

Basic metrics supported by a standard automatic differentiation libarary (ADL)

metric	definition
neural network
loss
(averaged) gradient

Advanced 1st/2nd-order metrics (FGHJK) supported by ASDL

metric	definition
Fisher information matrix
Fisher information matrix (MC estimation)
empirical Fisher
Gradient covariance
Hessian
Jacobian (per example)
Jacobian
Kernel

Matrix approximations

Supported operations

• matrix-vector product
  • power method
  • conjugate gradient method
• preconditioning gradient