[Draft] Concepts and Architecture #2
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jvmncs
reviewed
Oct 28, 2019
|
|
||
| - **Local tensors** are a specific type of TFE tensors used to represent plaintext values held locally by a player. Technically they are just small wrappers over raw tensors, but including extra metadata about locality and ownership. This metadata is used by e.g. kernels and protocols, as well as for detecting privacy violations. | ||
|
|
||
| - **Operations** are abstract targets used when defining secure computations, for instance `tfe.matmul` and `tfe.keras.Dense`. |
There was a problem hiding this comment.
Do we want to include tfe.keras.Dense as an Operation? A vanilla Dense contains several Operations, specifically MatMul and BiasAdd, but might not be an Operation itself.
|
|
||
| - **Protocols** are sets of kernels, and are hence not specific about *what* is being computed but rather *how* it is. They are typically used as context handlers, under which an abstract specification of a function is made concrete via the implied mapping from operations to kernels. Note that this roughly matches how "protocol" is used in e.g. MPC. | ||
|
|
||
| - **Functionalities** are protocols implementing a specific task and exposed as functions, for instance mapping from and to local tensors. This roughly follows the idea of ideal functionality from the UC framework yet are more intended to be user-facing. |
There was a problem hiding this comment.
What would the user want these for? I can see e.g. testing as one, but would be interested if you had others!
|
|
||
| - **Raw tensors** are the built-in TensorFlow tensors. | ||
|
|
||
| - **Local tensors** are a specific type of TFE tensors used to represent plaintext values held locally by a player. Technically they are just small wrappers over raw tensors, but including extra metadata about locality and ownership. This metadata is used by e.g. kernels and protocols, as well as for detecting privacy violations. |
There was a problem hiding this comment.
I'd also add representation as metadata (e.g. native TF data types, fixed-point, quantized, etc.)
mortendahl
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The aim of this RFC is to define a common vocabulary for TF Encrypted.