Comments (5)
This is a good start to the right idea, but I also feel it would add an effort to ensure this doesn’t break in future versions of SEAL implementation or even this library.
I would propose a separate helper tool to help define the appropriate encryption parameters and scale based off of the desired algorithmic depth. This could be thought of as the equivalent to the
CoeffModulus::BFVDefault
function used to generate the coeff moduli for the BFV scheme.This tool should not be coupled strongly to the native SEAL API nor this library any way, therefore posing no risk to either.
In TenSEAL, we assume that the user doesn't have enough experience in HE to deal with relin and rescaling, it would certainly be difficult on our side to deal with all these details while assuming general usage, but I think motivation(2) is a good assumption that simplify this solution. And of course we will need to give experienced user the ability to use different scale, for instance, users can still create CKKSVector with a different scale, but if they don't provide then we just use our context's scale. They should also be able to disable all these operations done after every multiplication if they want to deal with rescaling differently, but a rescaling strategy is really important for TenSEAL to be really simple to use.
from tenseal.
Good idea to simplify the API.
But since the scale is related to CKKS tensor, why don't we make it a CKKS tensor attribute, and generate or set it during tensor creation?
Mainly because of motivation(2), we don't need to set the scale for every tensor, all tensors should be having the same scale. The scale isn't a big field, using it in the TenSEALContext for CKKS and wasting that space for BFV isn't a big deal.
from tenseal.
In TenSEAL, we assume that the user doesn't have enough experience in HE to deal with relin and rescaling, it would certainly be difficult on our side to deal with all these details while assuming general usage, but I think motivation(2) is a good assumption that simplify this solution. And of course we will need to give experienced user the ability to use different scale, for instance, users can still create CKKSVector with a different scale, but if they don't provide then we just use our context's scale. They should also be able to disable all these operations done after every multiplication if they want to deal with rescaling differently, but a rescaling strategy is really important for TenSEAL to be really simple to use.
Okay, if you're already building a lot of helper logic then this makes sense to encapsulate as long as it is overridable.
from tenseal.
This is a good start to the right idea, but I also feel it would add an effort to ensure this doesn’t break in future versions of SEAL implementation or even this library.
I would propose a separate helper tool to help define the appropriate encryption parameters and scale based off of the desired algorithmic depth. This could be thought of as the equivalent to theCoeffModulus::BFVDefault
function used to generate the coeff moduli for the BFV scheme.
This tool should not be coupled strongly to the native SEAL API nor this library any way, therefore posing no risk to either.
from tenseal.
Good idea to simplify the API.
But since the scale is related to CKKS tensor, why don't we make it a CKKS tensor attribute, and generate or set it during tensor creation?
from tenseal.
Related Issues (20)
- Can't install tenseal using pip HOT 1
- Is gpu acceleration considered in future plans?
- Installation fails on macOS due to CMake error HOT 1
- (enc_vector1 + plain_vector1) not equal to enc(vector1 + vector1)
- Not getting installed using pip install tenseal
- Information conflict with SEAL regarding size of ciphertext under different key mode
- Failed to install TenSEAL on Macbook Air M2 HOT 1
- Multiparty extensions
- Tutorial 4 as a client/server scenario
- Defaulting to user installation because normal site-packages is not writeable Requirement already satisfied: tenseal in ./.local/lib/python3.10/site-packages (0.3.14)
- Failed to install tenseal, windows HOT 1
- Failed to install tenseal, windows HOT 2
- Rationale behind the parameter values in TenSEAL HOT 2
- How to rescale mannually?
- Windows 10 + python 3.8 ----- AttributeError: module 'tenseal' has no attribute 'Context' HOT 1
- Can't install TenSEAL on Windows
- Tenseal based federated learning: deep loop stuck and no longer executed in the case of loop encryption training parameters
- Why is my encrypted Linear Classifier slower than encrypted ConvNet classifier ?
- Is TenSEAL maintained?
- model training on encrypted data HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from tenseal.