SEDiL: Software for Edit Distance Learning

Metric learning constitutes a well-investigated field for vectorial data with successful applications, e.g. in computer vision, information retrieval, or bioinformatics. One particularly promising approach is offered by lowrank metric adaptation integrated into modern variants of learning vector quantization (LVQ). This technique is scalable with respect to both, data dimensionality and the number of data points, and it can be accompanied by strong guarantees of learning theory. Recent extensions of LVQ to general (dis-)similarity data have paved the way towards LVQ classifiers for non-vectorial, possibly discrete, structured objects such as sequences, which are addressed by classical alignment in bioinformatics applications. In this context, the choice of metric parameters plays a crucial role for the result, just as it does in the vectorial setting. In this contribution, we propose a metric learning scheme which allows for an autonomous learning of parameters (such as the underlying scoring matrix in sequence alignments) according to a given discriminative task in relational LVQ. Besides facilitating the often crucial and problematic choice of the scoring parameters in applications, this extension offers an increased interpretability of the results by pointing out structural invariances for the given task. *

show abstract

“…[12,34,3]. A popular platform which combines various adaptation methods for scoring functions is offered by SEDiL, for example [5].…”

Section: Motivation and Related Workmentioning

confidence: 99%

“…Replacement data: In this data set, all strings have 12 symbols, randomly generated from the alphabet Σ = {A, B, C, D} according to the regular expressions: (A|B) 5 (A|B) (C|D) (C|D) 5 for the first, and (A|B)…”

Section: Proof-of-concept With Artificial Datamentioning

confidence: 99%

Metric learning for sequences in relational LVQ

et al. 2015

View full text Add to dashboard Cite

show abstract

“…Each direction gives a di erent letter of the string (see Figure 2). We aim at comparing our approach to the one of [10], thanks to the use of SEDiL [11] and a weighted edit matrix. The matrix is learnt on the same data as (the input is the string from the learning set, the output the 1-nearest-neighbour).…”

Section: Handwritten Digit Classi Cationmentioning

confidence: 99%

Sequences Classification by Least General Generalisations

Tantini

Terlutte

Torre

2010

Grammatical Inference: Theoretical Results and Applications

View full text Add to dashboard Cite

In this paper, we present a general framework for supervised classi cation. This framework provides methods like boosting and only needs the de nition of a generalisation operator called lgg. For sequence classi cation tasks, lgg is a learner that only uses positive examples. We show that grammatical inference has already de ned such learners for automata classes like reversible automata or k-TSS automata. Then we propose a generalisation algorithm for the class of balls of words. Finally, we show through experiments that our method e ciently resolves sequence classi cation tasks.

show abstract

“…When it comes to sequence alignment in bioinformatics, lots of effort has been made for its correct choice based on biological insight [15]. Provided such insight is not always available, so-called inverse alignment can help to infer metric parameters from given, optimum alignments [16] [17]. In general, no such information is present, rather only weak learning signals such as motion labeling or grouping are available.…”

Section: Introductionmentioning

confidence: 99%

Efficient metric learning for the analysis of motion data

Hosseini¹,

Hammer²

2015

2015 IEEE International Conference on Data Science and Advanced Analytics (DSAA)

View full text Add to dashboard Cite

We investigate metric learning in the context of dynamic time warping (DTW), the by far most popular dissimilarity measure used for the comparison and analysis of motion capture data. While metric learning enables a problem-adapted representation of data, the majority of methods has been proposed for vectorial data only. In this contribution, we extend the popular principle offered by the large margin nearest neighbors learner (LMNN) to DTW by treating the resulting component-wise dissimilarity values as features. We demonstrate that this principle greatly enhances the classification accuracy in several benchmarks. Further, we show that recent auxiliary concepts such as metric regularization can be transferred from the vectorial case to component-wise DTW in a similar way. We illustrate that metric regularization constitutes a crucial prerequisite for the interpretation of the resulting relevance profiles.

show abstract

SEDiL: Software for Edit Distance Learning

Cited by 8 publications

References 9 publications

Metric learning for sequences in relational LVQ

Metric learning for sequences in relational LVQ

Sequences Classification by Least General Generalisations

Efficient metric learning for the analysis of motion data

Contact Info

Product

Resources

About