Abstract Concept Learning in Cognitive Robots

Current Robotics Reports ◽

10.1007/s43154-020-00038-x ◽

2021 ◽

Author(s):

Alessandro Di Nuovo ◽

Angelo Cangelosi

Keyword(s):

Concept Learning ◽

Cognitive Robotics ◽

Abstract Concept ◽

Abstract Concepts ◽

Knowledge Based ◽

Abstract Knowledge ◽

Benchmark Datasets ◽

Artificial Learning ◽

Concrete Concepts ◽

Cognitive Robots

Abstract Purpose of Review Understanding and manipulating abstract concepts is a fundamental characteristic of human intelligence that is currently missing in artificial agents. Without it, the ability of these robots to interact socially with humans while performing their tasks would be hindered. However, what is needed to empower our robots with such a capability? In this article, we discuss some recent attempts on cognitive robot modeling of these concepts underpinned by some neurophysiological principles. Recent Findings For advanced learning of abstract concepts, an artificial agent needs a (robotic) body, because abstract and concrete concepts are considered a continuum, and abstract concepts can be learned by linking them to concrete embodied perceptions. Pioneering studies provided valuable information about the simulation of artificial learning and demonstrated the value of the cognitive robotics approach to study aspects of abstract cognition. Summary There are a few successful examples of cognitive models of abstract knowledge based on connectionist and probabilistic modeling techniques. However, the modeling of abstract concept learning in robots is currently limited at narrow tasks. To make further progress, we argue that closer collaboration among multiple disciplines is required to share expertise and co-design future studies. Particularly important is to create and share benchmark datasets of human learning behavior.

Download Full-text

Situational systematicity: A role for schema in understanding the differences between abstract and concrete concepts

10.31234/osf.io/5qud8 ◽

2019 ◽

Cited By ~ 1

Author(s):

Charles P. Davis ◽

Gerry Altmann ◽

Eiling Yee

Keyword(s):

Concept Learning ◽

Brain Regions ◽

Control Network ◽

Abstract Concept ◽

Abstract Concepts ◽

Critical Determinant ◽

Per Se ◽

Concrete Concepts ◽

Insight Into ◽

Concrete Concept

Abstract concepts differ from concrete concepts in a number of ways. Here, we focus on what we refer to as situational systematicity: The objects and relations that constitute an abstract concept (e.g., justice) are more dispersed through space and time than are the objects and relations that typically constitute a concrete concept (e.g., chair); a larger set of objects and relations might potentially constitute an abstract concept than a concrete one; and exactly which objects and relations constitute a concept is likely more context-dependent for abstract than for concrete concepts. We thus refer to abstract concepts as having low situational systematicity. We contend that situational systematicity, rather than abstractness per se, may be a critical determinant of the cognitive, behavioral, and neural phenomena typically associated with concepts. We also contend that investigating concepts through the lens of schema provides insight into the situation-based dynamics of concept learning and representation, and into the functional significance of the interactions between brain regions that make up the schema control network.

Download Full-text

Exploring Geometric Feature Hyper-Space in Data to Learn Representations of Abstract Concepts

Applied Sciences ◽

10.3390/app10061994 ◽

2020 ◽

Vol 10 (6) ◽

pp. 1994 ◽

Cited By ~ 2

Author(s):

Rahul Sharma ◽

Bernardete Ribeiro ◽

Alexandre Miguel Pinto ◽

F. Amílcar Cardoso

Keyword(s):

Training Data ◽

Geometric Feature ◽

Automatic Identification ◽

Computational Domain ◽

Abstract Concept ◽

Abstract Concepts ◽

Concept Representation ◽

Data Problem ◽

Rans Modeling ◽

Concrete Concepts

The term concept has been a prominent part of investigations in psychology and neurobiology where, mostly, it is mathematically or theoretically represented. Concepts are also studied in the computational domain through their symbolic, distributed and hybrid representations. The majority of these approaches focused on addressing concrete concepts notion, but the view of the abstract concept is rarely explored. Moreover, most computational approaches have a predefined structure or configurations. The proposed method, Regulated Activation Network (RAN), has an evolving topology and learns representations of abstract concepts by exploiting the geometrical view of concepts, without supervision. In the article, first, a Toy-data problem was used to demonstrate the RANs modeling. Secondly, we demonstrate the liberty of concept identifier choice in RANs modeling and deep hierarchy generation using the IRIS dataset. Thirdly, data from the IoT’s human activity recognition problem is used to show automatic identification of alike classes as abstract concepts. The evaluation of RAN with eight UCI benchmarks and the comparisons with five Machine Learning models establishes the RANs credibility as a classifier. The classification operation also proved the RANs hypothesis of abstract concept representation. The experiments demonstrate the RANs ability to simulate psychological processes (like concept creation and learning) and carry out effective classification irrespective of training data size.

Download Full-text

Cognitive Learning Methodologies for Brain-Inspired Cognitive Robotics

International Journal of Cognitive Informatics and Natural Intelligence ◽

10.4018/ijcini.2015040103 ◽

2015 ◽

Vol 9 (2) ◽

pp. 37-54 ◽

Cited By ~ 27

Author(s):

Yingxu Wang

Keyword(s):

Cognitive Learning ◽

Knowledge Bases ◽

Cognitive Robotics ◽

Learning Theories ◽

Cognitive Mechanisms ◽

Knowledge Based ◽

Cognitive Knowledge ◽

Cognitive Foundations ◽

And Behavior ◽

Cognitive Robots

Cognitive robots are brain-inspired robots that are capable of inference, perception, and learning mimicking the cognitive mechanisms of the brain. Cognitive learning theories and methodologies for knowledge and behavior acquisition are centric in cognitive robotics. This paper explores the cognitive foundations and denotational mathematical means of cognitive learning engines (CLE) and cognitive knowledge bases (CKB) for cognitive robots. The architectures and functions of CLE are formally presented. A content-addressed knowledge base access methodology for CKB is rigorously elaborated. The CLE and CKB methodologies are not only designed to explain the mechanisms of human knowledge acquisition and learning, but also applied in the development of cognitive robots, cognitive computers, and knowledge-based systems.

Download Full-text

Abstract concepts and the suppression of arbitrary episodic context

10.31234/osf.io/nfy4s ◽

2019 ◽

Author(s):

Charles P. Davis ◽

Pedro M. Paz-Alonso ◽

Gerry Altmann ◽

Eiling Yee

Keyword(s):

Source Memory ◽

Abstract Concept ◽

Abstract Concepts ◽

Male And Female ◽

Female Voices ◽

Concrete Concepts ◽

Concrete Words

Context is important for abstract concept processing, but a mechanism by which it is encoded and re-instantiated with concepts is unclear. We used a source-memory paradigm to determine whether episodic context is attended more when processing abstract concepts. Experiment 1 presented abstract and concrete words in colored boxes at encoding. At test, memory for the frame color was worse for abstract concepts, counter to our predictions. Experiment 2 showed the same pattern when colored boxes were replaced with male and female voices. Experiment 3 presented words from encoding in the same or different box color to determine whether a greater advantage is conferred by context retention in memory for abstract concepts. There was instead a disadvantage: abstract concepts were less likely to be identified when the encoding color was retained at test. Concrete concepts are more sensitive to simple episodic detail, and in abstract concepts, arbitrary context may be suppressed.

Download Full-text

Exploring Geometric Feature Hyper-Space in Data to Learn Representations of Abstract Concepts

10.20944/preprints202001.0375.v1 ◽

2020 ◽

Author(s):

Rahul Sharma ◽

Bernardete Ribeiro ◽

Alexandre Miguel Pinto ◽

Amilcar F cardoso

Keyword(s):

Training Data ◽

Geometric Feature ◽

Automatic Identification ◽

Abstract Concept ◽

Abstract Concepts ◽

Concept Representation ◽

Concrete Concepts ◽

Hyper Space ◽

Hierarchy Generation ◽

Iris Data

The term Concept has been a prominent part of investigations in psychology and neurobiology where, mostly, it is mathematically or theoretically represented. The Concepts are also studied computationally through their symbolic, distributed and hybrid representations. The majority of these approaches focused on addressing concrete concepts notion, but the view of the abstract concept is rarely explored. Moreover, most computational approaches have a predefined structure or configurations. The proposed method, Regulated Activation Network (RAN), has an evolving topology and learns representations of Abstract Concepts by exploiting the geometrical view of Concepts, without supervision. In the article, the IRIS data was used to demonstrate: the RAN's modeling; flexibility in concept identifier choice; and deep hierarchy generation. Data from IoT's Human Activity Recognition problem is used to show automatic identification of alike classes as abstract concepts. The evaluation of RAN with 8 UCI benchmarks and the comparisons with 5 Machine Learning models establishes the RANs credibility as a classifier. The classification operation also proved the RAN's hypothesis of abstract concept representation. The experiments demonstrate the RANs ability to simulate psychological processes (like concept creation and learning) and carry out effective classification irrespective of training data size.

Download Full-text