Testing simple neuron models with dendrites for sparse binary image representation
This paper deals with the problem of information representation into a form that allows to make associations, measure similarity and integrate new information with respect to previously stored. Several simple models for encoding information into sparse distributed representation are explored. These...
Saved in:
| Date: | 2017 |
|---|---|
| Main Author: | |
| Format: | Article |
| Language: | English |
| Published: |
Інститут проблем штучного інтелекту МОН України та НАН України
2017
|
| Series: | Штучний інтелект |
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Journal Title: | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| Cite this: | Testing simple neuron models with dendrites for sparse binary image representation / V.M. Osaulenko // Штучний інтелект. — 2017. — № 2. — С. 101-108. — Бібліогр.: 21 назв. — англ. |
Institution
Digital Library of Periodicals of National Academy of Sciences of Ukraine| Summary: | This paper deals with the problem of information representation into a form that allows to make associations, measure similarity and integrate new information with respect to previously stored. Several simple models for encoding information into sparse distributed representation are explored. These models based on the idea that information about stimuli is stored in the population, not an individual neuron, thus each neuron learns many partial features. Results show formation of a sparse representation of image data with high overlap for similar images. Each cell develops multiple receptive fields that together create a population receptive field. It was possible due to incorporation of dendritic tree into standard neuron model. Also, models were tested on a classification of handwritten digits from MNIST dataset. Results from unsupervised representation show poor accuracy compared to the state-of-the-art supervised methods, however, due to the presence of interesting properties further development of an idea should be continued. |
|---|