Andrea Di Ferdinando

Evolving Modular Architectures for Neural Networks

Neural networks that learn the What and Where task perform better if they possess a modular architecture for separately processing the identity and spatial location of objects. In previous simulations the modular architecture either was hardwired or it developed during an individual’s life based on a preference for short connections given a set of hardwired unit locations. We present two sets of simulations in which the network architecture is genetically inherited and it evolves in a population of neural networks in two different conditions: (1) both the architecture and the connection weights evolve; (2) the network architecture is inherited and it evolves but the connection weights are learned during life. The best results are obtained in condition (2). Condition (1) gives unsatisfactory results because (a) adapted sets of weights can suddenly become maladaptive if the architecture changes, (b) evolution fails to properly assign computational resources (hidden units) to the two tasks, (c) genetic linkage between sets of weights for different modules can result in a favourable mutation in one set of weights being accompanied by an unfavourable mutation in another set of weights.

Modeling Orienting Behavior and Its Disorders with “Ecological” Neural Networks

Computational modeling is a useful tool for spelling out hypotheses in cognitive neuroscience and testing their predictions in artificial systems. Here we describe a series of simulations involving neural networks that learned to perform their task by self-organizing their internal connections. The networks controlled artificial agents with an orienting eye and an arm. Agents saw objects with various shapes and locations and learned to press a key appropriate to their shape. The results showed the following: (1) Despite being able to see the entire visual scene without moving their eye, agents learned to orient their eye toward a peripherally presented object. (2) Neural networks whose hidden layers were previously partitioned into units dedicated to eye orienting and units dedicated to arm movements learned the identification task faster and more accurately than did nonmodular networks. (3) Nonetheless, even nonmodular networks developed a similar functional segregation through self-organization of their hidden layer. (4) After partial disconnection of the hidden layer from the input layer, the lesioned agents continued to respond accurately to single stimuli, wherever they occurred, but on double simultaneous stimulation they oriented toward and responded only to the right-sided stimulus, thus simulating extinction/neglect. These results stress the generality of the advantages provided by orienting processes. Hard-wired modularity, reminiscent of the distinct cortical visual streams in the primate brain, provided further evolutionary advantages. Finally, disconnection is likely to be a mechanism of primary importance in the pathogenesis of neglect and extinction symptoms, consistent with recent evidence from animal studies and brain-damaged patients.

BRIDGING DIGITAL AND PHYSICAL EDUCATIONAL GAMES USING RFID/NFC TECHNOLOGIES

The physical educational games and the traditional psycho-pedagogical methodologies are deeply based on the manipulation of objects. The opportunity derived by some low-cost technologies could join the physical world with the digital tools creating Augmented Reality Environments based on the concepts of Internet of Things (IoT). This connection has all the capacities to enhance the traditional educational games played in the schools or at home with the digital tools in order to create more exiting learning activities and more appropriate for the new digital natives. In this field, the RFID/NFC technology seems to be a natural candidate due to its natural predisposition to be heavily connected to real objects and send the signal to the digital devices. In this paper, we describe how the RFID/NFC technology could be used to connect digital and physical didactic materials in this hybrid approach. We present three different applications and prototypes: a) Block-Magic, it is an educational games based on well-known Logic Blocks material in the framework of an European project (I); b) Walden PECS Communicator (II) a platform based on Picture Exchange Communication System (PECS) a worldwide methodology to enhance communication skills in autistic persons; c) WandBot (III, IV), a learning environment that combines toy robots, RFID-technology and serious games for scientific dissemination in science centres

Block Magic: A Prototype Bridging Digital and Physical Educational Materials to Support Children Learning Processes

Block Magic is a prototype for educational materials developed in a successful European research project under the framework of LLP-Comenius programme. It aimed at creating a bridge between physical manipulation and digital technology in education. Block Magic developed a functional prototypal system that enhanced the Logic Blocks Box. The prototype is made up of an active desk/board able to recognise concrete blocks equipped with the RFID passive tag and to communicate with a PC, an augmented reality system. Preliminary trials with Block Magic prototype were run in various schools in Germany, Greece, Italy and Spain, involving children aged 3 to 7. Results confirmed Block Magic educational platform effectiveness in educational context.

Objects, spatial compatibility, and affordances: A connectionist study

In two Artificial Life simulations we evolved artificial organisms possessing a visual and a motor system, and whose nervous system was simulated with a neural network. Each organism could see four objects, either upright or reversed, with a left or a right handle. In Task 1 they learned to reach the object handle independently of the handles position. In Task 2 they learned to reach one of two buttons located below the handle either to decide where the handle was (Simulation 1) or whether the object was upright or reversed (Simulation 2). Task 1 simulated real life experience, Task 2 replicated either a classic spatial compatibility task (Simulation 1) or an experiment by Tucker and Ellis (1998) (Simulation 2). In both simulations learning occurred earlier in the Compatible condition, when the button to reach and the handle were on the same side, than in the Incompatible condition.

Towards Hyper Activity Books for Children. Connecting Activity Books and Montessori-like Educational Materials

In the first years of school, activity books and Montessori-like educational materials are widespread. They satisfy children’s precise psychological needs and result in funded educational practices based on game activity adopted by teachers. These materials are more effective for promoting learning if used in close interaction with an adult, they cannot be customised and the corrective feedback cannot be provided within the appropriate time frame. In this paper, we aim to overcome these limitations by exploiting a Technology Enhanced Learning (TEL) methodology to connect activity books and Montessori-like educational materials. We propose a general architecture for building hyper activity and exercise books called HAB, with three levels: multimedia, multimodality and computing; we describe a first implementation for validating this architecture, Block Magic, and outline development of the architecture under the INF@NZIA DIGI.Tales 3.6 project.

Visuospatial planning in the travelling salesperson problem: a connectionist account of normal and impaired performance.

Planning is a fundamental cognitive function frequently employed in common daily activities. The Travelling Salesperson Problem (TSP), in which participants decide what order between a number of locations optimizes total travel distance, is a paradigm that allows the study of planning and strategy choice. In the TSP, subjects adopt visuo-spatial heuristics to perform the task and operate a continuous monitoring to adapt their behaviour. We present a connectionist model of the TSP that simulates bottom-up and top-down influences observed in the execution of the task. The model accounts for the continuous monitoring observed in healthy participants, and, after a simulated lesion, it also accounts for the decrease of heuristic switching observed in frontal patients and in normal subjects under repetitive transcranial magnetic stimulation (rTMS) over frontal lobe.

Action and hierarchical levels of categories: A connectionist perspective

Recent views of categorization suggest that categories are action-based rather than arbitrary symbols. Three connectionist simulations explore the hierarchical organization of categories in the framework of an action-based theory of categorization. In the simulations an organism with a visual system and a two-segment arm has to reach different points in space depending on the object seen and on context. The context indicates whether to put the object in a superordinate or in a basic category. The results show that: (a) superordinate categories are easier to learn than basic ones; (b) the more similar the actions to perform with basic and superordinate categories, the easier to learn the task; (c) violation of category boundaries leads to less good performance.

Tangible Interfaces for Cognitive Assessment and Training in Children: LogicART

This paper describes how to use tangible interfaces for cognitive assessment and training. Assessment and training are two fundamental phases in the learning process: assessment allows to gather information in order to identify the learner starting level and monitor progresses that training can lead to. Assessment uses tests, some of which are addressed to cognitive abilities. Many tests for cognitive abilities are based on verbal materials that are not suitable for population such as young children or people with special needs, and people with hard cognitive disabilities or sensory impairment. In these cases it would be more fruitable to use physical objects. This paper illustrates a hardware-software system, LogicART, that can be used in assessment and training of cognitive abilities, such as reasoning, memory, categorization, etc. exploiting physical materials augmented by RFID/NFC Technology. These materials, employed to study cognitive abilities, offer the advantage to stimulate multisensoriality and manipulation.

Grounding Serious Game Design on Scientific Findings: The Case of ENACT on Soft Skills Training and Assessment

The lack of open-access tool for the enhancement and promotion of soft skills is bringing the e-learning community to new educational challenges. The paper describes the implementation of ENACT, an online serious game for the standardised psychometric assessment and training of users’ negotiation skills through the interaction with virtual artificial agents. The assessment process is divided into 8 scenarios based on real life situations and investigates the user negotiation styles in relation to Rahim’s conceptualization of five different styles of handling conflict. Need analysis data and preliminary testing results of the platform are presented.