tensorflow / tensorflow/tensorboard

Faster, more performant t-SNE

Open
#731 5 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

plugin:projector stat:contributions welcome theme:performance
Dominant language
TypeScript
Stars
7.2k
Forks
1.7k
Avg merge
4d 22h
Merged PRs (30d)
1

Description

Hi,

I am working on unsupervised clustering featuers vectors (length of 2048) of large sets of images (40k to 200k). I would like to do this using t-SNE and then explore it using the Tensorboard embedding visualizer. But the performance on Tensorboard is to poor for the task. This is probably because it is run on client side and I can only apply one cpu/gpu for it.
It would be nice if I could do t-SNE clustering and run a 1000 iterations offline (and use multi threaded processing) and then load the result in the Tensorboard embedding visualizer so I can explore and label it.

Looking forward to hear your ideas on this.

Greetings,
Nicolaas

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

The issue names no files, tests, or entry points. Start by reviewing TensorBoard's current client-side t-SNE flow and embedding visualizer, then determine how offline, 1000-iteration processing and multithreaded execution could integrate. Done would mean large image-vector sets can be processed offline and the results loaded for exploration and labeling.

Written by the indexing model from the issue text.

Assessment

Domain
data-visualization, machine-learning, performance
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.