https://github.com/namin/relaxed-machines
program synthesis with neuro-symbolic differentiable interpreters
https://github.com/namin/relaxed-machines
differentiable-programming interpreters jax neuro-symbolic paper-implementations synthesis
Last synced: 4 months ago
JSON representation
program synthesis with neuro-symbolic differentiable interpreters
- Host: GitHub
- URL: https://github.com/namin/relaxed-machines
- Owner: namin
- Created: 2021-12-30T23:26:26.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2024-01-09T14:20:08.000Z (over 1 year ago)
- Last Synced: 2025-02-09T01:44:25.667Z (4 months ago)
- Topics: differentiable-programming, interpreters, jax, neuro-symbolic, paper-implementations, synthesis
- Language: Python
- Homepage:
- Size: 149 KB
- Stars: 10
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
Relaxed Machines
----------------
Program synthesis in neuro-symbolic differentiable interpreters.
## Inspired by ∂4
Differential interpreters in which the weights of the machine can be interpreted as a program.
I think this is different in spirit from ∂4, which corresponds to hard sketches here with holes that are whole neural networks.
Baby steps (see the [log](TODO.md#log)):
1. [`inc_stop`](inc_stop.py)
2. [`dup_add`](dup_add.py)
3. [`reg_jmp`](reg_jmp.py)
4. [`sub`](sub.py)
- [Notes](NOTES.md)
- [Run logs](LOG.md)
## Libraries
- [JAX](https://github.com/google/jax)
- [Haiku](https://github.com/deepmind/dm-haiku)
- [Optax](https://github.com/deepmind/optax)
## Resources
- Chapter 3 (∂4: A Differentiable Forth Interpreter) of Matko Bošnjak's Ph.D thesis, [On Differentiable Interpreters](https://discovery.ucl.ac.uk/id/eprint/10121772/), UCL, 2021.
## Acks
Many thanks to [Matko Bošnjak](https://matko.info/) and [Rob Zinkov](https://zinkov.com) for discussions, insights, suggestions, and pointers.