awslabs / awslabs/aws-cv-task2vec

what would a task2vec (cosine) distance of -1 mean?

Open
#14 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
126
Forks
31
PR merge metrics
No merged PRs in 30d

Description

I feel -1 would mean the task representations/embeddings point in opposite directions. But the actual vectors are the same. So are the tasks completely similar up to a single bit of info (the sign) or is it better to interpret this that the two tasks are the same and the direction should be ignored?

In general I'm trying to figure how to interpret the negative sign.

I suppose 0 means perpendicular tasks and 1 means the actual same vector -- even the sign.

Contributor guide

Open the contributing guide

Research direction

Start with the Task2Vec paper linked in the repository description and inspect how cosine distance is used in this project. Compare the mathematical definition with the reported vector behavior, then document the intended interpretation of negative values; done when the sign convention and similarity meaning are clearly explained.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Documentation
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.