Poster
Logical Languages Accepted by Transformer Encoders with Hard Attention
Pablo Barcelo · Alexander Kozachinskiy · Anthony W. Lin · Vladimir Podolskii
Halle B #195
Abstract:
We contribute to the study of formal languages that can be recognized by transformer encoders. We focus on two self-attention mechanisms: (1) UHAT (Unique Hard Attention Transformers) and (2) AHAT (Average Hard Attention Transformers). UHAT encoders are known to recognize only languages inside the circuit complexity class AC0, i.e., accepted by a family of poly-sized and depth-bounded boolean circuits with unbounded fan-ins. On the other hand, AHAT encoders can recognize languages outside AC0), but their expressive power still lies within the bigger circuit complexity class TC0, i.e., AC0-circuits extended by majority gates.We first show a negative result that there is an AC0-language that cannot be recognized by an UHAT encoder. On the positive side, we show that UHAT encoders can recognize a rich fragment of AC0-languages, namely, all languages definable in first-order logic with arbitrary unary numerical predicates. This logic, includes, for example, all regular languages from AC0. We then show that AHAT encoders can recognize all languages of our logic even when we enrich it with counting terms. Using these results, we obtain a characterization of which counting properties are expressible by UHAT and AHAT, in relation to regular languages.
Chat is not available.