gamaleldin Goto Github PK

cfr

This code package is for the Corrected-Fisher-Randomization (CFR) method. This method generates random surrogate data that preserves a specified set of first and second order marginal moments of a data tensor, which makes it well equipped to test for the null hypothesis that a structure in data is an epiphenomenon of these specified set of primary features of the data tensor. The randomization procedure used in CFR is based on Fisher randomization (shuffling). However, the shuffling is accompanied by a correction step that retains the primary features specified in the null hypothesis. Hence, the name of this method.

cleverhans

An adversarial example library for constructing attacks, building defenses, and benchmarking both

code-challenge

ff_generator_model

Feed-Forward Generator Model

google-research

Google Research

maxvar_subspaces

Code for optimizing max variance subspaces from Elsayed & Lara et al. Nature Communications 2016

models

Models and examples built with TensorFlow

playground

rand_subspaces

Code for sampling random subspaces from Elsayed & Lara et al. Nature Communications 2016

rand_tensor

This code package generates random tensors with user specified marginal means and covariances from one of two optional distributions. The first is the maximum-entropy-distribution with the specified marginal means and covariances. The second is the tensor-normal-distribution with the specified means and covariances.

retinalresources

Code for "A Unified Theory of Early Visual Representations from Retina to Cortex through Anatomically Constrained Deep CNNs", ICLR 2019

tensorflow

Computation using data flow graphs for scalable machine learning

tme

This code package is for the Tensor-Maximum-Entropy (TME) method. This method generates random surrogate data that preserves a specified set of first and second order marginal moments of a data tensor, which makes it well equipped to test for the null hypothesis that a structure in data is an epiphenomenon of these specified set of primary features of the data tensor. The random surrogate data are sampled from a maximum entropy distribution. This distribution unlike traditional maximum entropy method have constraints on the marginal first and second moments of the tensor mode.