Encoding Reusable Perceptual Features Enables Learning Future Categories from Few Examples
نویسندگان
چکیده
A perceptual system coping with a dynamic environment must be able to learn to detect new object categories from a few examples. However, learning from a small sample is restricted by the hindering effects of model overfitting. We present an algorithm aimed at circumventing the effects of overfitting by utilizing a set of reusable features, learned from several previously trained categories. We show that when applying a feature reuse strategy the algorithm learns complex real world objects from few examples.
منابع مشابه
Representational Shifts during Category Acquisition: A Preference for Features that Provide Information for Multiple Categories
The present study was aimed at examining the factors inducing feature creation in the perceptual and semantic systems. The proposed research was guided by the hypothesis that due to the substantial load of categorization tasks, a preference exists for encoding features that are informative for the recognition of multiple categories. This claim extends theories stipulating that features are enco...
متن کاملPreferential Encoding of Features Distinctive for Multiple Categories
Understanding how features are encoded during category acquisition is a fundamental challenge in the study of human learning. The current work proposes that in order to maintain accurate generalization in large scale categorization systems, features that are useful in discriminating multiple categories must be actively preferred. This multi-class hypothesis stands in contrast with theories in w...
متن کاملBiologically inspired Bayesian approach for learning object categories from few training examples
In this work we present a biologically inspired algorithm for learning object categories that uses Bayesian inference to integrate information within and across fixations. In our model, an object is represented as a collection of features of specific classes arranged at specific locations with respect to the location of the fixation point. Even though the number of feature detectors that we use...
متن کاملThe role of attention allocation during induction
Induction is an essential aspect of human learning and reasoning, yet there is no consensus regarding the underlying mechanisms. One view holds, early in development, induction is similarity-based, utilizing perceptual features, possibly allowing increased encoding therefore higher memory accuracy. While another view posits that across development induction requires identifying category members...
متن کاملThe Effect of Semantic Mapping as a Vocabulary Instruction Technique on EFL Learners with Different Perceptual Learning Styles
Traditional and modern vocabulary instruction techniques have been introduced in the past few decades to improve the learners’ performance in reading comprehension. Semantic mapping, which entails drawing learners’ attention to the interrelationships among lexical items through graphic organizers, is claimed to enhance vocabulary learning significantly. However, whether this technique suits all...
متن کامل