| EPSRC Reference: |
GR/J07228/01 |
| Title: |
LOGICAL NEURAL SYSTEMS FOR LANGUAGE, IMAGE AND ACTION ASSOCIATION. |
| Principal Investigator: |
Aleksander, Professor I |
| Other Investigators: |
|
| Researcher Co-Investigators: |
|
| Project Partners: |
|
| Department: |
Electrical and Electronic Engineering |
| Organisation: |
Imperial College London |
| Scheme: |
Standard Research (Pre-FEC) |
| Starts: |
01 January 1994 |
Ends: |
31 December 1995 |
Value (£): |
92,406
|
| EPSRC Research Topic Classifications: |
|
| EPSRC Industrial Sector Classifications: |
|
| Related Grants: |
|
| Panel History: |
|
|
Summary on Grant Application Form |
|
To develop and study the feasibility of a general logical neural system operating in a near-real-world environment. The system will relate language-like symbol strings to internal iconic representations, and respond to environmental input with appropriate language-like output.Progress:The following deliverables were specified to be produced by the investigating team at Imperial College.Months 0-6Addition of an output field capability to the MAGNUS weightless neural network simulator, developed as part of grant GR/F34756. This has been achieved, allowing the system to generate outputs in several forms which can specify the sensed position of a simulated system within the coordinate system of its near-real-world environment.Testing systems on MAGNUS with differing iconic resolutions.Several different networks have been simulated using MAGNUS: it has proved to be capable of supporting relatively large iconic resolutions of up to about 40 feedback inputs per input field, depending on the size of input fields. Work by a PhD student in the research group, Antonio D.P. Braga, has produced results on storage capacity and probabilities for pattern contradictions during storage. These empirical and theoretical results allow informed configuration of simulated systems for optimal performance. Development of the simple word interface.Further development of the MAGNUS simulation software has allowed input to simulated networks to be partitioned into separate visual, linguistic, and motor-sensory modalities. This partitioning has allowed the investigation of several representation systems for input words: alphabetic,1-in-N bit encodings, and visual text images. These systems together form the basis of the word interface developed. Learning of the world image.MAGNUS has been extended to allow simulation of networks and interaction in real time with grey-scale video input from the KITCHENWORLD (near real-world) environment. Learning of both linguistic symbols for specific visual objects, and also classification of unseen objects in the same categories, has been demonstrated. A paper has been submitted to the Cooperative Multimodal Communication conference, CMC-95, on this topic.
|
|
Key Findings |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
|
Potential use in non-academic contexts |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
|
Impacts |
|
| Description |
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk |
| Summary |
|
| Date Materialised |
|
|
|
|
Sectors submitted by the Researcher |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
| Project URL: |
|
| Further Information: |
|
| Organisation Website: |
http://www.imperial.ac.uk |