<p>Artificial intelligence is a broad research area and there are many different reasons why it is interesting to study artificial intelligence. One of the main reasons is to understand how information might be represented in the human brain. The Recursive Auto Associative Memory (RAAM) is a connectionist architecture that with some success has been used for that purpose since it develops compact distributed representations for compositional structures.</p><p>A lot of extensions to the RAAM architecture have been developed through the years in order to improve the performance of RAAM; Bi coded RAAM (B-RAAM) is one of those extensions. In this work a modified B-RAAM architecture is tested and compared to RAAM regarding: Training speed, ability to learn with smaller internal representations and generalization ability. The internal representations of the two network models are also analyzed and compared. This dissertation also includes a discussion of some theoretical aspects of B-RAAM.</p><p>It is found here that the training speed for B-RAAM is considerably lower than RAAM, on the other hand, RAAM learns better with smaller internal representations and is better at generalize than B-RAAM. It is also shown that the extracted internal representation of RAAM reveals more structural information than it does for B-RAAM. This has been shown by hieratically cluster the internal representation and analyse the tree structure. In addition to this a discussion is added about the justifiability to label B-RAAM as an extension to RAAM.</p>

<p>Connectionist models have been criticized for not being able to form compositional representations of recursive data structures such as trees and lists, a matter that has been addressed by models as Elman networks, RAAM and B-RAAM. These architectures seem to have common features with the human short-term memory regarding recall. Both show a strong recency effect; however, the human memory also exhibits a primacy effect due to rehearsal. The problem is that the connectionist models do not have the primacy aspect, which complicates the learning of long-term dependencies. A long-term dependency is when items presented early should affect the behaviour of the model. Learning long-term dependencies is a problem that is hard to address within these architectures.</p><p>Delay-lines might be used as a mechanism for implementing rehearsal within connectionist models. However, it has not been clarified how the use of delay-lines affects the recency and the primacy aspect. In this thesis, delay-lines are introduced in B-RAAM. This study investigates how the primacy and the recency aspect are affected by the use of delay-lines, aiming to improve the ability to identify long-term dependencies. The results show that by using delay-lines, B-RAAM has both primacy and recency.</p>

Connectionist models have been criticized for not being able to form compositional representations of recursive data structures such as trees and lists, a matter that has been addressed by models as Elman networks, RAAM and B-RAAM. These architectures seem to have common features with the human short-term memory regarding recall. Both show a strong recency effect; however, the human memory also exhibits a primacy effect due to rehearsal. The problem is that the connectionist models do not have the primacy aspect, which complicates the learning of long-term dependencies. A long-term dependency is when items presented early should affect the behaviour of the model. Learning long-term dependencies is a problem that is hard to address within these architectures. Delay-lines might be used as a mechanism for implementing rehearsal within connectionist models. However, it has not been clarified how the use of delay-lines affects the recency and the primacy aspect. In this thesis, delay-lines are introduced in B-RAAM. This study investigates how the primacy and the recency aspect are affected by the use of delay-lines, aiming to improve the ability to identify long-term dependencies. The results show that by using delay-lines, B-RAAM has both primacy and recency.

Artificial intelligence is a broad research area and there are many different reasons why it is interesting to study artificial intelligence. One of the main reasons is to understand how information might be represented in the human brain. The Recursive Auto Associative Memory (RAAM) is a connectionist architecture that with some success has been used for that purpose since it develops compact distributed representations for compositional structures. A lot of extensions to the RAAM architecture have been developed through the years in order to improve the performance of RAAM; Bi coded RAAM (B-RAAM) is one of those extensions. In this work a modified B-RAAM architecture is tested and compared to RAAM regarding: Training speed, ability to learn with smaller internal representations and generalization ability. The internal representations of the two network models are also analyzed and compared. This dissertation also includes a discussion of some theoretical aspects of B-RAAM. It is found here that the training speed for B-RAAM is considerably lower than RAAM, on the other hand, RAAM learns better with smaller internal representations and is better at generalize than B-RAAM. It is also shown that the extracted internal representation of RAAM reveals more structural information than it does for B-RAAM. This has been shown by hieratically cluster the internal representation and analyse the tree structure. In addition to this a discussion is added about the justifiability to label B-RAAM as an extension to RAAM.