Shared-Knowledge-Lifelong-Learnin
Shared-Knowledge-Lifelong-Learning
| Shared-Knowledge-Lifelong-Learnin | Shared-Knowledge-Lifelong-Learning | |
|---|---|---|
| 1 | 1 | |
| - | 24 | |
| - | - | |
| - | 6.6 | |
| - | 6 months ago | |
| Python | ||
| - | Creative Commons Zero v1.0 Universal |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
Shared-Knowledge-Lifelong-Learnin
-
AI That Teaches Other AI
> Their SKILL tool involves a set of algorithms that make the process go much faster, they said, because the agents learn at the same time in parallel. Their research showed if 102 agents each learn one task and then share, the amount of time needed is reduced by a factor of 101.5 after accounting for the necessary communications and knowledge consolidation among agents.
This is a really interesting idea. It's like the reverse of knowledge distillation (which I've been thinking about a lot[1]) where you have one giant model that knows a lot about a lot & you use that model to train smaller, faster models that know a lot about a little.
Instead, you if you could train a lot of models that know a lot about a little (which is a lot less computationally intensive because the problem space is so confined) and combine them into a generalized model, that'd be hugely beneficial.
Unfortunately, after a bit of digging into the paper & Github repo[2], this doesn't seem to be what's happening at all.
> The code will learn 102 small and separte heads(either a linear head or a linear head with a task bias) for each tasks respectively in order. This step can be parallized on multiple GPUS with one task per GPU. The heads will be saved in the weight folder. After that, the code will learn a task mapper(Either using GMMC or Mahalanobis) to distinguish image task-wisely. Then, all images will be evaluated in the same time without a task label.
So the knowledge isn't being combined (and the agents aren't learning from each other) into a generalized model. They're just training a bunch of independent models for specific tasks & adding a model-selection step that maps an image to the most relevant "expert". My guess is you could do the same thing using CLIP vectors as the routing method to supervised models trained on specific datasets (we found that datasets largely live in distinct regions of CLIP-space[3]).
[1] https://github.com/autodistill/autodistill
[2] https://github.com/gyhandy/Shared-Knowledge-Lifelong-Learnin...
[3] https://www.rf100.org
Shared-Knowledge-Lifelong-Learning
-
AI That Teaches Other AI
> Their SKILL tool involves a set of algorithms that make the process go much faster, they said, because the agents learn at the same time in parallel. Their research showed if 102 agents each learn one task and then share, the amount of time needed is reduced by a factor of 101.5 after accounting for the necessary communications and knowledge consolidation among agents.
This is a really interesting idea. It's like the reverse of knowledge distillation (which I've been thinking about a lot[1]) where you have one giant model that knows a lot about a lot & you use that model to train smaller, faster models that know a lot about a little.
Instead, you if you could train a lot of models that know a lot about a little (which is a lot less computationally intensive because the problem space is so confined) and combine them into a generalized model, that'd be hugely beneficial.
Unfortunately, after a bit of digging into the paper & Github repo[2], this doesn't seem to be what's happening at all.
> The code will learn 102 small and separte heads(either a linear head or a linear head with a task bias) for each tasks respectively in order. This step can be parallized on multiple GPUS with one task per GPU. The heads will be saved in the weight folder. After that, the code will learn a task mapper(Either using GMMC or Mahalanobis) to distinguish image task-wisely. Then, all images will be evaluated in the same time without a task label.
So the knowledge isn't being combined (and the agents aren't learning from each other) into a generalized model. They're just training a bunch of independent models for specific tasks & adding a model-selection step that maps an image to the most relevant "expert". My guess is you could do the same thing using CLIP vectors as the routing method to supervised models trained on specific datasets (we found that datasets largely live in distinct regions of CLIP-space[3]).
[1] https://github.com/autodistill/autodistill
[2] https://github.com/gyhandy/Shared-Knowledge-Lifelong-Learnin...
[3] https://www.rf100.org
What are some alternatives?
autodistill - Images to inference with no labeling (use foundation models to train supervised models).
roboflow-100-benchmark - Code for replicating Roboflow 100 benchmark results and programmatically downloading benchmark datasets