Recurrent Neural Network (RNN)
Definition
RNNs consider previous inputs through a passed internal state.
A neural network that passes an internal state through a sequence, thereby considering earlier inputs.
Simply explained
When reading a sequence, an intermediate state is passed along so that previous information is not immediately lost.
Why it matters in practice
The term helps to select, develop, evaluate AI systems in a well-founded manner and realistically assess their technical risks.
Example from business
An RNN processes sensor measurements in temporal order.
Limits and pitfalls
- Expecting very long dependencies, ignoring gradient problems, or not masking padding.
Origin & history
Recurrent networks were established for sequences in the 1980s and 1990s; LSTM later addressed long-term dependencies.
Use in mid-sized companies
- Possible example: An RNN processes sensor measurements in temporal order.
Frequently asked questions
What does recurrent neural network (RNN) mean?
A neural network that passes an internal state through a sequence, thereby considering earlier inputs. What mistakes should be avoided? Expecting very long dependencies, ignoring gradient problems, or not masking padding.
Sources
- Finding Structure in Time. Cognitive Science, 14(2), 179–211. · Jeffrey L. Elman (1990): Finding Structure in Time. Cognitive Science, 14(2), 179–211.
- Frei zugängliche veröffentlichte Fassung · Frei zugängliche veröffentlichte Fassung
Editorially reviewed on: