To enable comparisons with other one-shot classification models, one could try to perform one-shot SOM learning. This sounds difficult, but there is some literature on the matter which we could at least try and evaluate. Since the model has some cognitive plausibility and all that, I think that just some competitive performance (not really state of the art, but close) could be interesting.
To enable comparisons with other one-shot classification models, one could try to perform one-shot SOM learning. This sounds difficult, but there is some literature on the matter which we could at least try and evaluate. Since the model has some cognitive plausibility and all that, I think that just some competitive performance (not really state of the art, but close) could be interesting.
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.70.8986&rep=rep1&type=pdf https://drum.lib.umd.edu/handle/1903/1746