A biologically motivated visual memory architecture for online learning of objects

Abstract

We present a biologically motivated architecture for object recognition that is based on a hierarchical feature detection model in combination with a memory architecture that implements short-term and long-term memory for objects. A particular focus is the functional realization of online and incremental learning for the task of appearance-based object recognition of many complex-shaped objects. We propose some modifications of learning vector quantization algorithms that are especially adapted to the task of incremental learning and capable of dealing with the stabilityplasticity dilemma of such learning algorithms. Our technical implementation of the neural architecture is capable of online learning of 50 objects within less than three hours.

Combining offline and online classifiers for life-long learning

“…We use the learning technique proposed in [23] that has gained much attention recently in the context of big data and interpretable models due to its flexibility and intuitive classification scheme, see e. g. [33], [34], [35], [36], [37], [38], [39], [40], [41], [42]. Essentially, it offers an efficient way for prototypebased data classification.…”

Towards autonomous bootstrapping for life-long learning categorization tasks

“…The final number of allocated nodes w k and the assigned category labels u k corresponds to the difficulty of the different categories itself but also to the within-category variance. Finally the long-term stability of these incrementally learned nodes is considered based on an individual node learning rate Θ k as proposed in [7].…”

“…Similar to Step 1 we test new LVQ nodes only for erroneous categories. In contrast to the node insertion rule proposed in [7], where nodes are inserted for training vectors with smallest distance to wrong winning nodes, we propose to insert new LVQ nodes based on training vectors x i with most categorization errors. This leads to a more compact representation, because a single node typically improves the representation of several categories.…”

A biologically inspired approach for interactive learning of categories

“…Thus stability can be better achieved compared to the multi-layer perceptron (MLP), where all weights are modified at each learning step. Additionally for life-long learning architectures often a node specific learning rate combined with an incremental node insertion rule (Hamker, 2001;Furao & Hasegawa, 2006;Kirstein et al, 2008) is used to approach the "stability plasticity dilemma". The major drawback of those identification architectures is that they inefficiently separate co-occurring classes.…”