Online research in philosophy

PDP networks that use nonmonotonic activation functions often produce hidden unit regularities that permit the internal structure of these networks to be interpreted (Berkeley et al., 1995; McCaughan, 1997; Dawson, 1998). In particular, when the responses of hidden units to a set of patterns are graphed using jittered density plots, these plots organize themselves into a set of discrete stripes or bands. In some cases, each band is associated with a local interpretation. On the basis of these observations, Berkeley (2000) has suggested that these bands are both subsymbolic and symbolic in nature, and has used the analysis of one network to support the claim that there are fewer differences between symbols and subsymbols than one might expect. We suggest below that this conclusion is premature. First, in many cases the local interpretation of each band is difficult to relate to the interpretation of a network's response; a more appropriate relationship only emerges when a band associated with one hidden unit is considered in the context of other bands associated with other hidden units (i.e., interpretations of distributed representations are more useful than interpretations of local representations). Second, the content that a band designates to an external observer (i.e., the interpretation assigned to a band by the researcher) can be quite different from the content that a band designates to the output units of the network itself.. We use two different network simulations – including the one described by Berkeley (2000) – to illustrate these points. We conclude that current evidence involving interpretations of nonmonotonic PDP networks actually illustrates the differences between symbolic and subsymbolic processing.

The employment of a particular class of computer programs known as "connectionist networks" to model mental processes is a widespread approach to research in cognitive science these days. Little has been written, however, on the precise connection that is thought to hold between such programs and actual in vivo cognitive processes such that the former can be said to "model" the latter in a scientific sense. What is more, this relation can be shown to be problematic. In this paper I give a brief overview of the use of connectionist models in cognitive science, and then explore some of the statements connectionists have made about the nature of the "modeling relation" thought to hold between them and cognitive processes. Finally I show that these accounts are inadequate and that more work is necessary if connectionist networks are to be seriously regarded as scientific models of cognitive processes.

This paper examines whether a classical model could be translated into a PDP network using a standard connectionist training technique called extra output learning. In Study 1, standard machine learning techniques were used to create a decision tree that could be used to classify 8124 different mushrooms as being edible or poisonous on the basis of 21 different Features (Schlimmer, 1987). In Study 2, extra output learning was used to insert this decision tree into a PDP network being trained on the identical problem. An interpretation of the trained network revealed a perfect mapping from its internal structure to the decision tree, representing a precise translation of the classical theory to the connectionist model. In Study 3, a second network was trained on the mushroom problem without using extra output learning. An interpretation of this second network revealed a different algorithm for solving the mushroom problem, demonstrating that the Study 2 network was indeed a proper theory translation.

Classical cognitive science assumes that intelligently behaving systems must be symbol processors that are implemented in physical systems such as brains or digital computers. By contrast, connectionists suppose that symbol manipulating systems could be approximations of neural networks dynamics. Both classicists and connectionists argue that symbolic computation and subsymbolic dynamics are incompatible, though on different grounds. While classicists say that connectionist architectures and symbol processors are either incompatible or the former are mere implementations of the latter, connectionists reply that neural networks might be incompatible with symbol processors because the latter cannot be implementations of the former. In this contribution, the notions of 'incompatibility' and 'implementation' will be criticized to show that they must be revised in the context of the dynamical system approach to cognitive science. Examples for implementations of symbol processors that are incompatible with respect to contextual topologies will be discussed.

This paper offers both a theoretical and an experimental perspective on the relationship between connectionist and Classical (symbol-processing) models. Firstly, a serious flaw in Fodor and Pylyshyn’s argument against connectionism is pointed out: if, in fact, a part of their argument is valid, then it establishes a conclusion quite different from that which they intend, a conclusion which is demonstrably false. The source of this flaw is traced to an underestimation of the differences between localist and distributed representation. It has been claimed that distributed representations cannot support systematic operations, or that if they can, then they will be mere implementations of traditional ideas. This paper presents experimental evidence against this conclusion: distributed representations can be used to support direct structure-sensitive operations, in a man- ner quite unlike the Classical approach. Finally, it is argued that even if Fodor and Pylyshyn’s argument that connectionist models of compositionality must be mere implementations were correct, then this would still not be a serious argument against connectionism as a theory of mind.

Recently, connectionist models have been developed that seem to exhibit structuresensitive cognitive capacities without executing a program. This paper examines one such model and argues that it does execute a program. The argument proceeds by showing that what is essential to running a program is preserving the functional structure of the program. It has generally been assumed that this can only be done by systems possessing a certain temporalcausal organization. However, counterfactualpreserving functional architecture can be instantiated in other ways, for example geometrically, which are realizable by connectionist networks.

This paper critically examines the claim that parallel distributed processing (PDP) networks are autonomous learning systems. A PDP model of a simple distributed associative memory is considered. It is shown that the 'generic' PDP architecture cannot implement the computations required by this memory system without the aid of external control. In other words, the model is not autonomous. Two specific problems are highlighted: (i) simultaneous learning and recall are not permitted to occur as would be required of an autonomous system; (ii) connections between processing units cannot simultaneously represent current and previous network activation as would be required if learning is to occur. Similar problems exist for more sophisticated networks constructed from the generic PDP architecture. We argue that this is because these models are not adequately constrained by the properties of the functional architecture assumed by PDP modelers. It is also argued that without such constraints, PDP researchers cannot claim to have developed an architecture radically different from that proposed by the Classical approach in cognitive science.

This paper examines the use of connectionism (neural networks) in modelling legal reasoning. I discuss how the implementations of neural networks have failed to account for legal theoretical perspectives on adjudication. I criticise the use of neural networks in law, not because connectionism is inherently unsuitable in law, but rather because it has been done so poorly to date. The paper reviews a number of legal theories which provide a grounding for the use of neural networks in law. It then examines some implementations undertaken in law and criticises their legal theoretical naïvete. It then presents a lessons from the implementations which researchers must bear in mind if they wish to build neural networks which are justified by legal theories.

This paper presents considerations in favour of the view that traditional (classical) architectures can be seen as emergent features of connectionist networks with distributed representation. A recent paper by William Bechtel (1988) which argues for a similar conclusion is unsatisfactory in that it fails to consider whether the compositional syntax and semantics attributed to mental representations by classical models can emerge within a connectionist network. The compatibility of the two paradigms hinges largely, I suggest, on how this question is answered. Focusing on the issue of syntax, I argue that while such structure is lacking in connectionist models with local representation, it can be accommodated within networks where representation is distributed. I discuss an important paper by Smolenski (1988) which attempts to show how connectionists can incorporate the relevant syntactic structure, suggesting that some criticisms levelled against that paper by Fodor & Pylyshyn (1988) are wanting. I then go on to indicate a strategy by which a compositional syntax and semantics can be defined for the sort of network that Smolenski describes. I conclude that since the connectionist can respect the central tenets of classicism, the two approaches are compatible with one another.

Connectionism is a style of modeling based upon networks of interconnected simple processing devices. This style of modeling goes by a number of other names too. Connectionist models are also sometimes referred to as 'Parallel Distributed Processing' (or PDP for short) models or networks.1 Connectionist systems are also sometimes referred to as 'neural networks' (abbreviated to NNs) or 'artificial neural networks' (abbreviated to ANNs). Although there may be some rhetorical appeal to this neural nomenclature, it is in fact misleading as connectionist networks are commonly significantly dissimilar to neurological systems. For this reason, I will avoid using this terminology, other than in direct quotations. Instead, I will follow the practice I have adopted above and use 'connectionist' as my primary term for systems of this kind.