google-deepmind/dnc

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

2531 stars · 446 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A TensorFlow implementation of the Differentiable Neural Computer (DNC), a recurrent neural network architecture with an external memory published in Nature by DeepMind. It is built from composable RNNCore modules (controller, access, temporal linkage, freeness) and includes an example training script for algorithmic tasks.

Use cases

• implement a differentiable neural computer in tensorflow
• experiment with memory-augmented neural network architectures
• train an RNN on sequence copying tasks
• study the DNC architecture from the Nature paper
• build custom memory access modules for RNNs

When to choose

• you want a faithful reference implementation of the DNC paper
• you work in TensorFlow/Sonnet and want pluggable RNNCore memory modules
• you're doing research on memory-augmented neural networks

When to avoid

• you use PyTorch or JAX instead of TensorFlow
• you need actively maintained code with recent updates
• you want a production-ready sequence model rather than research code

Facets

library · maturity maintenance

machine-learning deep-learning deep-learning machine-learning python tensorflow sonnet recurrent-neural-network external-memory research-code deepmind algorithms

Member repositories

For agents

markdown · JSON · MCP: product_card(name="google-deepmind/dnc")