A Pattern-Hierarchy Classifier for Reduced Teaching

Abstract

This paper describes a design that can be used for Explainable AI. The lower level is a nested ensemble of patterns created by self-organisation. The upper level is a hierarchical tree, where nodes are linked through individual concepts, so there is a transition from mixed ensemble masses to specific categories. Lower-level pattern ensembles are learned in an unsupervised manner and then split into branches when it is clear that the category has changed. Links between the two levels define that the concepts are learned and missing links define that they are guessed only. This paper proposes some new clustering algorithms for producing the pattern ensembles, that are themselves an ensemble which converges through aggregations. Multiple solutions are also combined, to make the final result more robust. One measure of success is how coherent these ensembles are, which means that every data row in the cluster belongs to the same category. The total number of clusters is also important and the teaching phase can correct the ensemble estimates with respect to both of these. A teaching phase would then help the classifier to learn the true category for each input row. During this phase, any classifier can learn or infer correct classifications from some other classifier's knowledge, thereby reducing the required number of presentations. As the information is added, cross-referencing between the two structures allows it to be used more widely, where a unique structure can build up that would not be possible by either method separately.

New ideas for brain modelling 3

“…Because this paper is a consolidation of earlier work, any of the author's papers [3]- [13] are of interest. Some other work is also mentioned, to show that a lot of the established theories are also relevant.…”

“…A separate paper [3] has suggested a self-organising algorithm that uses a full-linking closestnodes part first, followed by the Frequency Grid [7]. The algorithm therefore also uses an ensemble at the base, with a hierarchy above that, but the hierarchy is for agglomerative clustering.…”

“…That would probably encourage Hebbian growth in the mini-cluster only, especially if distance and synchronization are related. These units that aggregate individual nodes can then be aggregated themselves in a third level, when we get the self-organising algorithm described in [3]. The algorithm produces something that is thus like a unit of work and maybe a more complex representation of the ReN [9].…”

“…This paper proposes that the memory structures can be formed in a largely unsupervised manner, in the sense that they represent and organise data using limited but automatic processes. The structure can maybe then be corrected when it links with the experiencebased level and interacts with a different type of knowledge [3]. Rote learning is therefore still part of the system, but 'units of work' are created first through self-organisation, before they can be further clustered into knowledge trees, for example, using rote learning.…”

A 2-Level Memory Architecture for Brain Modelling

“…Rather than the program recognising associated antonyms, or something like that, it may have combined 2 lower clusters somewhere, but the algorithms in this paper are different to what was used previously. The problem for a self-organising system was to some extent solved in [7] through averaging over multiple solutions, but that would also be a future project.…”