TensorFlow/Differential Privacy - Revision history

Admin: /* dp_sgd Building */

2017-10-27T03:04:40Z

dp_sgd Building

← Older revision		Revision as of 03:04, 27 October 2017
Line 68:		Line 68:
	Build instructions do not work...		Build instructions do not work...

			See issue 2605: https://github.com/tensorflow/models/issues/2605

	=Flags=		=Flags=

Admin: /* dp_sgd Building */

2017-10-27T03:03:36Z

dp_sgd Building

← Older revision		Revision as of 03:03, 27 October 2017
Line 67:		Line 67:

	Build instructions do not work...		Build instructions do not work...



			=Flags=

			[[Category:NN]]
			[[Category:ML]]
			[[Category:TensorFlow]]

Admin: /* Prerequisites */

2017-10-27T03:02:54Z

Prerequisites

← Older revision		Revision as of 03:02, 27 October 2017
Line 33:		Line 33:
	* per_example_gradients		* per_example_gradients

	====Prerequisites====		====dp_sgd Prerequisites====

	Preparation steps, covered individually below:		Preparation steps, covered individually below:
Line 63:		Line 63:
	mnist_train.tfrecord		mnist_train.tfrecord
	</pre>		</pre>

			====dp_sgd Building====

			Build instructions do not work...

Admin: /* dp_sgd subdirect */

2017-10-27T01:26:21Z

dp_sgd subdirect

← Older revision		Revision as of 01:26, 27 October 2017
Line 32:		Line 32:
	* dp_optimizer		* dp_optimizer
	* per_example_gradients		* per_example_gradients

			====Prerequisites====

			Preparation steps, covered individually below:
			* Install TensorFlow (old version)
			* Install Bezel
			* Prepare slim version of MNIST dataset

			Installing old version of Tensorflow:

			<pre>
			$ virtualenv --system-site-packages tf
			$ source tf/bin/activate
			(tf)$ easy_install -U pip
			(tf)$ pip install tensorflow==0.12.1
			</pre>

			Installing Bezel: using 0.5.4 (used Homebrew to install...)

			Preparing slim MNIST data set:

			<pre>
			$ cd models/research/slim/

			$ python download_and_convert_data.py --dataset_name=mnist --dataset_dir="${DATA_DIR}"

			$ ls ${DATA_DIR}
			labels.txt
			mnist_test.tfrecord
			mnist_train.tfrecord
			</pre>

Admin: /* Directories and Components */

2017-10-27T01:08:35Z

Directories and Components

← Older revision		Revision as of 01:08, 27 October 2017
Line 20:		Line 20:

	The top level directory contains three directories:		The top level directory contains three directories:
	* dp_sgd		* dp_sgd - develops algorithmic techniques for learning and a refined analysis of privacy costs within the framework of differential privacy (implementation of Deep Learning with Differential Privacy paper)
	* multiple_teachers		* multiple_teachers - creates student and teacher networks, transfer knowledge from teacher networks to student networks in a differentially private manner by noisily aggregating the teacher decisions before feeding them to students (implementation of Semi-Supervised Knowledge Transfer paper)
	* privacy_accountant		* privacy_accountant - not sure yet...

			===dp_sgd subdirect===

			Directory is focused

			This directory itself contains subdirectories:
			* dp_mnist
			* dp_optimizer
			* per_example_gradients

Admin: Created page with "=Overview= The differential privacy network is a component in the [https://github.com/tensorflow/models tensorflow/models] repository on Github. Link to code: https://github..."

2017-10-27T01:04:25Z

Created page with "=Overview= The differential privacy network is a component in the [https://github.com/tensorflow/models tensorflow/models] repository on Github. Link to code: https://github..."

New page

=Overview=

The differential privacy network is a component in the [https://github.com/tensorflow/models tensorflow/models] repository on Github.

Link to code: https://github.com/tensorflow/models/tree/master/research/differential_privacy

The networks are based on two papers:

"Deep Learning with Differential Privacy" (October 2016)

Link to paper: https://arxiv.org/pdf/1607.00133.pdf

"Semi-supervised Knowledge Transfer for Deep Learning from Private Training Data" (March 2017)

Link to paper: https://arxiv.org/pdf/1610.05755.pdf

This repository directory is quite messy, so I'm trying to clean it up a bit.

==Directories and Components==

The top level directory contains three directories:
* dp_sgd
* multiple_teachers
* privacy_accountant