Shimon Edelman

Learning to recognize visual objects from examples requires the ability to find meaningful patterns in spaces of very high dimensionality. We present a method for dimensionality reduction which effectively biases the learning system by combining multiple constraints via the use of class labels. The use of extensive class labels steers the resulting lowdimensional representation to become invariant to those directions of variation in the input space that are irrelevant to classification; this is done merely by making class labels independent of these directions. We also show that prior knowledge of the proper dimensionality of the target representation can be imposed by training a multi-layer bottleneck network. Computational experiments involving non-trivial categorization of parameterized fractal images and of human faces indicate that the low-dimensional representation extracted by our method leads to improved generalization in the learned tasks and is likely to preserve the topology of the original space

A computational theory of consciousness should include a quantitative measure of consciousness, or MoC, that (i) would reveal to what extent a given system is conscious, (ii) would make it possible to compare not only different systems, but also the same system at different times, and (iii) would be graded, because so is consciousness. However, unless its design is properly constrained, such an MoC gives rise to what we call the boundary problem: an MoC that labels a system as conscious will do so for some – perhaps most – of its subsystems, as well as for irrelevantly extended systems (e.g., the original system augmented with physical appendages that contribute nothing to the properties supposedly supporting consciousness), and for aggregates of individually conscious systems (e.g., groups of people). This problem suggests that the properties that are being measured are epiphenomenal to consciousness, or else it implies a bizarre proliferation of minds. We propose that a solution to the boundary problem can be found by identifying properties that are intrinsic or systemic: properties that clearly differentiate between systems whose existence is a matter of fact, as opposed to those whose existence is a matter of interpretation (in the eye of the beholder). We argue that if a putative MoC can be shown to be systemic, this ipso facto resolves any associated boundary issues. As test cases, we analyze two recent theories of consciousness in light of our definitions: the Integrated Information Theory and the Geometric Theory of consciousness.

The proponents of machine consciousness predicate the mental life of a machine, if any, exclusively on its formal, organizational structure, rather than on its physical composition. Given that matter is organized on a range of levels in time and space, this generic stance must be further constrained by a principled choice of levels on which the posited structure is supposed to reside. Indeed, not only must the formal structure fit well the physical system that realizes it, but it must do so in a manner that is determined by the system itself, simply because the mental life of a machine cannot be up to an external observer. To illustrate just how tall this order is, we carefully analyze the scenario in which a digital computer simulates a network of neurons. We show that the formal correspondence between the two systems thereby established is at best partial, and, furthermore, that it is fundamentally incapable of realizing both some of the essential properties of actual neuronal systems and some of the fundamental properties of experience. Our analysis suggests that, if machine consciousness is at all possible, conscious experience can only be instantiated in a class of machines that are entirely different from digital computers, namely, time-continuous, open, analog dynamical systems.

A standing challenge for the science of mind is to account for the datum that every mind faces in the most immediate – that is, unmediated – fashion: its phenomenal experience. The complementary tasks of explaining what it means for a system to give rise to experience and what constitutes the content of experience (qualia) in computational terms are particularly challenging, given the multiple realizability of computation. In this paper, we identify a set of conditions that a computational theory must satisfy for it to constitute not just a sufficient but a necessary, and therefore naturalistic and intrinsic, explanation of qualia. We show that a common assumption behind many neurocomputational theories of the mind, according to which mind states can be formalized solely in terms of instantaneous vectors of activities of representational units such as neurons, does not meet the requisite conditions, in part because it relies on inactive units to shape presently experienced qualia and implies a homogeneous representation space, which is devoid of intrinsic structure. We then sketch a naturalistic computational theory of qualia, which posits that experience is realized by dynamical activity-space trajectories (rather than points) and that its richness is measured by the representational capacity of the trajectory space in which it unfolds.

Scientific theories of consciousness identify its contents with the spatiotemporal structure of neural population activity. We follow up on this approach by stating and motivating Dynamical Emergence Theory, which defines the amount and structure of experience in terms of the intrinsic topology and geometry of a physical system’s collective dynamics. Specifically, we posit that distinct perceptual states correspond to coarse-grained macrostates reflecting an optimal partitioning of the system’s state space—a notion that aligns with several ideas and results from computational neuroscience and cognitive psychology. We relate DET to existing work, offer predictions for empirical studies, and outline future research directions.

We introduce a set of biologically and computationally motivated design choices for modeling the learning of language, or of other types of sequential, hierarchically structured experience and behavior, and describe an implemented system that conforms to these choices and is capable of unsupervised learning from raw natural‐language corpora. Given a stream of linguistic input, our model incrementally learns a grammar that captures its statistical patterns, which can then be used to parse or generate new data. The grammar constructed in this manner takes the form of a directed weighted graph, whose nodes are recursively (hierarchically) defined patterns over the elements of the input stream. We evaluated the model in seventeen experiments, grouped into five studies, which examined, respectively, (a) the generative ability of grammar learned from a corpus of natural language, (b) the characteristics of the learned representation, (c) sequence segmentation and chunking, (d) artificial grammar learning, and (e) certain types of structure dependence. The model's performance largely vindicates our design choices, suggesting that progress in modeling language acquisition can be made on a broad front—ranging from issues of generativity to the replication of human experimental findings—by bringing biological and computational considerations, as well as lessons from prior efforts, to bear on the modeling approach.

To the extent that the performance of embodied and situated cognitive agents is predicated on fore- thought;such agents must remember; and learn from; the past to predict the future. In complex; non-stationaryenvironments; such learning is facilitated by an intrinsic motivation to seek novelty. A significant part of anagent’s identity is thus constituted by its remembered distilled cumulative life experience; which the agent isdriven to constantly expand. The combination of the drive to novelty with practical limits on memorycapacity posits a problem. On the one hand; because novelty seekers are unhappy when bored; merelyreliving past positive experiences soon loses its appeal: happiness can only be attained sporadically; via anopen-ended pursuit of new experience. On the other hand; because the experiencer’s memory is finite;longevity and continued novelty; taken together; imply eventual loss of at least some of the stored content;and with it a disruption of the constructed identity. In this essay; I examine the biobehavioral and cognitive-computational circumstances that give rise to this problem and explore its implications for the humancondition.

Construction-based approaches to syntax (Croft, 2001; Goldberg, 2003) posit a lexicon populated by units of various sizes, as envisaged by (Langacker, 1987). Constructions may be specified completely, as in the case of simple morphemes or idioms such as take it to the bank, or partially, as in the expression what’s X doing Y?, where X and Y are slots that admit fillers of particular types (Kay and Fillmore, 1999). Constructions offer an intriguing alternative to traditional rule-based syntax by hinting at the extent to which the complexity of language can stem from a rich repertoire of stored, more or less entrenched (Harris, 1998) representations that address both syntactic and semantic issues, and encompass, in addition to general rules, “totally idiosyncratic forms and patterns of all intermediate degrees of generality” (Langacker, 1987, p.46). Because constructions are by their very nature language-specific, the question of acquisition in Construction Grammar is especially poignant. We address this issue by offering an unsupervised algorithm that learns constructions from raw corpora.

We examined the role of fitness, commonly assumed without proof to be conferred by the mastery of language, in shaping the dynamics of language evolution. To that end, we introduced island migration (a concept borrowed from population genetics) into the shared lexicon model of communication (Nowak et al., 1999). The effect of fitness linear in language coherence was compared to a control condition of neutral drift. We found that in the neutral condition (no coherence-dependent fitness) even a small migration rate – less than 1% – suffices for one language to become dominant, albeit after a long time. In comparison, when fitness-based selection is introduced, the subpopulations stabilize quite rapidly to form several distinct languages. Our findings support the notion that language confers increased fitness. The possibility that a shared language evolved as a result of neutral drift appears less likely, unless migration rates over evolutionary times were extremely small.

Converging evidence from anatomical studies (Maunsell, 1983) and functional analyses (Hubel & Wisesel, 1968) of the nervous system suggests that the feed-forward pathway of the mammalian perceptual system follows a largely hierarchic organization scheme. This may be because hierarchic structures are intrinsically more viable and thus more likely to evolve (Simon, 2002). But it may also be because objects in our environment have a hierarchic structure and the perceptual system has evolved to match it. We conducted a behavioral experiment to investigate the effect of the degree of hierarchy of the generative probabilistic structure in categorization. We generated one set of stimuli using a hierarchic underlying probability distribution, and another set according to a non-hierarchic one. Participants were instructed to categorize these images into one of the two possible categories a. Our results suggest that participants perform more accurately in the case of hierarchically structured stimuli.

We compare our model of unsupervised learning of linguistic structures, ADIOS [1, 2, 3], to some recent work in computational linguistics and in grammar theory. Our approach resembles the Construction Grammar in its general philosophy (e.g., in its reliance on structural generalizations rather than on syntax projected by the lexicon, as in the current generative theories), and the Tree Adjoining Grammar in its computational characteristics (e.g., in its apparent affinity with Mildly Context Sensitive Languages). The representations learned by our algorithm are truly emergent from the (unannotated) corpus data, whereas those found in published works on cognitive and construction grammars and on TAGs are hand-tailored. Thus, our results complement and extend both the computational and the more linguistically oriented research into language acquisition. We conclude by suggesting how empirical and formal study of language can be best integrated.

Variation set structure — partial alignment of successive utterances in child-directed speech — has been shown to correlate with progress in the acquisition of syntax by children. The present study demonstrates that arranging a certain proportion of utterances in a training corpus in variation sets facilitates word segmentation and phrase structure learning in miniature artifi- cial languages by adults. Our findings have implications for understanding the mechanisms of L1 acquisition by children, and for the development of more efficient algorithms for automatic language acquisition, as well as better methods for L2 instruction.

A theory of representation is incomplete if it states “representations areX” whereXcan be symbols, cell assemblies, functional states, or the flock of birds fromTheaetetus, without explaining the nature of the link between the universe ofXs and the world. Amit's thesis, equating representations with reverberations in Hebbian cell assemblies, will only be considered a solution to the problem of representation when it is complemented by a theory of how a reverberation in the brain can be a representation of anything.

To learn a visual code in an unsupervised manner, one may attempt to capture those features of the stimulus set that would contribute significantly to a statistically efficient representation. Paradoxically, all the candidate features in this approach need to be known before statistics over them can be computed. This paradox may be circumvented by confining the repertoire of candidate features to actual scene fragments, which resemble the “what+where” receptive fields found in the ventral visual stream in primates. We describe a single-layer network that learns such fragments from unsegmented raw images of structured objects. The learning method combines fast imprinting in the feedforward stream with lateral interactions to achieve single-epoch unsupervised acquisition of spatially localized features that can support systematic treatment of structured objects [1]

We report a quantitative analysis of the cross-utterance coordination observed in child-directed language, where successive utterances often overlap in a manner that makes their constituent structure more prominent, and describe the application of a recently published unsupervised algorithm for grammar induction to the largest available corpus of such language, producing a grammar capable of accepting and generating novel wellformed sentences. We also introduce a new corpus-based method for assessing the precision and recall of an automatically acquired generative grammar without recourse to human judgment. The present work sets the stage for the eventual development of more powerful unsupervised algorithms for language acquisition, which would make use of the coordination structures present in natural child-directed speech.