Diss. Ph. D.--Columbus--Ohio state university, 1973. / Bibliogr. p. 337-346.

This dissertation modifies error-correcting codes and Bloom filters to create high-capacity associative memories. These associative memories use principled statistical inference and are distributed as no single component contains enough information to complete the task by itself, yet the components can collectively solve the task by passing information to each other. These associative memories are also robust to hardware failure as their distributed nature ensures there is no single point of failure. This dissertation starts by simplifying a Bloom filter so that it tolerates hardware failure (albeit with reduced performance). An efficient associative memory is created by performing inference over the set of items stored in the Bloom filter. This architecture suggests a modification which forgets old patterns stored in the associative memory (known as a palimpsest memory). It is shown that overwriting old patterns in an independent manner reduces performance, but is still comparable to the well-known Hopfield network. The lost performance can be regained using integer storage which allows the superposition of the pattern representation, or ensuring bits are not overwritten independently using concepts from errorcorrecting codes. The final task performs recall in continuous time using components which are more similar to neurons than used in the rest of the dissertation. The resulting memory has the exciting ability to recall many patterns simultaneously. Statistical inference ensures gradual degradation of the performance as an associative memory is overloaded. Since many definitions of associative memory capacity rely on the existence of catastrophic failure a new definition of capacity is provided. In spite of some biologically unrealistic attributes, this work is relevant to the understanding of the brain as it provides high performance solutions to the associative memory task which is known to be relevant to the brain.

No description available.

The contextual cueing effect (CCE) is a phenomenon that shows that our brains can take advantage of invariant contextual information in our environment to help us locate targets or relevant information more efficiently. In a seminal study by Chun and Jiang (1998), participants searched for a target letter “T” among “L” distractors. Unbeknownst to the participants, some trials had repeated configurations, while others had novel ones. Participants found the “T” faster in repeated configurations, showing implicit learning. Classical studies demonstrated learning of only single context-target pairing. However, recent research (Wang et al., 2020) shows that learning could also happen for repeated contexts paired with one of multiple (e.g., 4) target locations. In the current study, we intended to examine such learning at the individual scene level by producing matching target eccentricity between a pair of repeated and novel scenes. We varied the magnitude of four target eccentricities by producing equal spacing (in Experiment 1) or variable spacing (in Experiments 2 and 3) of both repeated and novel scenes. Experiment 1 showed comparable learning for different target locations with different eccentricities except for targets with the smallest eccentricity. In Experiment 2, we compared conditions with targets concentrated on the larger versus smaller eccentricity range in a between-subject design, and we found that at least when the target appeared in a large eccentricity, CCE was larger when the target appeared in the distribution condition with larger eccentricity bias than distribution with low eccentricity bias. However, this trend appeared present even in the classical contextual cueing paradigm with one target paired with one repeated context. In Experiment 3, we performed the same manipulation of eccentricity distribution in the classical contextual cueing paradigm and found the effect seen in Experiment 2 was not robust. These results suggest that when a given target could be paired with multiple repeated contexts, the learning of target-context association is more flexible and can be modulated by the target's location probability. / Thesis / Master of Science (MSc) / Through our daily interactions with the environment, we learn consistent relations between objects. For example, in a classroom with a fixed seating arrangement, the teacher can find a particular student with greater ease after repeated exposures. This type of learning can be demonstrated in laboratory settings through the contextual cueing effect (CCE). When participants encounter repeated search displays, they locate targets faster than if the display was novel. This learning occurs implicitly. Even when the repeated context is paired with multiple possible target locations learning still occurs (Wang et al., 2020). The present studies investigate this multiple target location–context pairing using a paradigm similar to that used by Wang et al.. We ask whether each target location is learned equally well when multiple targets are paired with repeated contexts, and explore if location probability cueing influences the CCE. Our results suggest that participants can learn multiple target locations equally well and may also learn the target's location probability. Humans interact with their environment daily, learning consistent relationships between objects. For example, in a classroom with a fixed seating arrangement, finding a particular student becomes easier. This type of learning is known as the contextual cueing effect (CCE) in laboratory settings. When participants encounter repeated search displays, they locate targets faster than in novel ones, and this learning occurs implicitly. Recent findings (Wang et al., 2020) show that learning also occurs when a repeated context is paired with one of multiple possible targets, as long as those targets pair with other repeated contexts. The goals of the current studies are to investigate, in the paradigm used by Wang et al. 2020, whether each target location is learned equally well when multiple targets are paired with repeated contexts and to determine if location probability cueing can influence CCE. Our results suggest that participants can learn target locations equally well and may also learn the target's location probability.

Notre étude est centrée sur les mécanismes elliptiques au sein des anaphores associativesméronymiques. Nous sommes partie de l’hypothèse que dans ce type d’anaphore, il existe deuxstructures : une structure profonde et une structure de surface. La première consiste en la présence destrois éléments : le tout, le prédicat partitif et la partie. La deuxième, où apparaissent les différents typesd’ellipse, fait l’objet de notre travail.Nous nous sommes attardée sur trois types d’ellipse que nous considérons caractéristiques desanaphores méronymiques : l’ellipse du prédicat partitif, celle du deuxième élément de la structure N DeN et celle de l’antécédent anaphorique. Traitées séparément, les anaphores nominales, les anaphoresverbales et les anaphores adverbiales ont été soumises dans un premier temps à une descriptionsyntactico-sémantique et dans un deuxième temps à la théorie des trois fonctions primaires. Cettethéorie nous a permis d’expliquer la possibilité pour certains éléments, d’être élidés au sein del’anaphore associative méronymique. / Our study focuses on elliptic mechanisms within associative meronymic anaphora. We departed fromthe assumption that in this type of anaphora, there are two structures: a deep structure and a surfacestructure. The first consists in the presence of three elements: the whole, the partitive predicate and thepart. The second, which shows the different types of ellipsis, is the topic of our work.We focused on three types of Ellipsis, which we considered typical meronymic anaphora: ellipsis ofpartitive predicate, ellipsis of the second element of the structure [N De N] and that of the anaphoricantecedent. Treated separately, nominal anaphora, verbal anaphora, and adverbial anaphora wereinitially submitted to a syntactic-semantic description, and then, to the three primary functions theory.This theory allowed us to explain the ability of certain items to be elided in the associative meronymicanaphora.

No description available.

No description available.

No description available.

No description available.

No description available.