Fixed Point vs. First-Order Logic on Finite Ordered Structures with Unary Relations

We prove that first order logic is strictly weaker than fixed point logic over every infinite classes of finite ordered structures with unary relations: Over these classes there is always an inductive unary relation which cannot be defined by a first-order formula, even when every inductive sentence (i.e., closed formula) can be expressed in first-order over this particular class. Our proof first establishes a property valid for every unary relation definable by first-order logic over these classes which is peculiar to classes of ordered structures with unary relations. In a second step we show that this property itself can be expressed in fixed point logic and can be used to construct a non-elementary unary relation.

An Infinite Pebble Game and Applications

We generalize the well-known pebble game to infinite dag's, and we use this generalization to give new and shorter proofs of results in different areas of computer science (as diverse as "logic of programs" and "formal language theory"). Our applications here include a proof of a theorem due to Salomaa, asserting the existence of a context-free language with infinite index, and a proof of a theorem due to Tiuryn and Erimbetov, asserting that unbounded memory increases the power of logics of programs. The original proofs by Salomaa, Tiuryn, and Erimbetov, are fairly technical. The proofs by Tiuryn and Erimbetov also involve advanced techniques of model theory, namely, back-and-forth constructions based on a variant of Ehrenfeucht-Fraisse games. By contrast, our proofs are not only shorter, but also elementary. All we need is essentially finite induction and, in the case of the Tiuryn-Erimbetov result, the compactness and completeness of first-order logic.

Self-Organizing Grammar Induction Using a Neural Network Model

This paper presents a self-organizing, real-time, hierarchical neural network model of sequential processing, and shows how it can be used to induce recognition codes corresponding to word categories and elementary grammatical structures. The model, first introduced in Mannes (1992), learns to recognize, store, and recall sequences of unitized patterns in a stable manner, either using short-term memory alone, or using long-term memory weights. Memory capacity is only limited by the number of nodes provided. Sequences are mapped to unitized patterns, making the model suitable for hierarchical operation. By using multiple modules arranged in a hierarchy and a simple mapping between output of lower levels and the input of higher levels, the induction of codes representing word category and simple phrase structures is an emergent property of the model. Simulation results are reported to illustrate this behavior.