Cogito componentiter ergo sum
Summary (1 min read)
1 Introduction
- During evolution human and animal visual, auditory, and other primary sensory systems have adapted to a broad ecological ensemble of natural stimuli.
- It is a fascinating finding in many real world data sets that the label structure discovered by unsupervised learning closely coincides with labels obtained by letting a human or a group of humans perform classification, labels derived from human cognition.
- A term set is chosen and a document is represented by the vector of term frequencies.
- LSI/PCA was then performed on the sparsified feature coefficients for visualization.
- The results seem to indicate that generalizable cognitive components corresponding to the phoneme /ae/ opening the utterances s and f, can be identified using linear component analysis.
3 Conclusion
- Cognitive component analysis (COCA) has been defined as the process of unsupervised grouping of data such that the ensuing group structure is well-aligned with that resulting from human cognitive activity.
- It is well-established that information theoretically optimal representations, similar to those found by ICA, are in use in several information processing tasks in human and animal perception.
- By visualization of data using latent semantic analysis-like plots, the authors have shown that independent components analysis is also relevant for representing semantic structure, in text and other abstract data such as social networks, musical features, and speech.
- The authors therefore speculate that the cognitive machinery developed for analyzing complex perceptual signals from multi-agent environments may also be used in higher brain function.
- Hence, the authors hypothesize that independent component analysis –given the right representation– may be a quite generic tool for COCA.
Did you find this useful? Give us your feedback
Citations
1 citations
Cites result from "Cogito componentiter ergo sum"
...This indicates that core elements of lyrical music appear to be treated in a fashion similar to those of language [18], which is in turn supported by EEG `electroencephalograhy' studies showing that language and music compete for the same neural resources when processing syntax and semantics [19]....
[...]
1 citations
Cites background from "Cogito componentiter ergo sum"
...[24, 8], has shown that the spatial receptive field properties match sparse representations, and by maximizing statistical independence and sparseness of the representations, the resulting receptive field properties share similarities with those of cortical neurons [66]....
[...]
...They anticipated the predictive power of abstract unsupervised learning techniques [8]....
[...]
Cites methods from "Cogito componentiter ergo sum"
...Thus far, ICA has been used to model the ray structure and to represent the semantic structure in text, the communities in social networks, and other abstract data, e.g. music (Hansen et al., 2005; Hansen & Feng, 2006) and speech (Feng & Hansen, 2006)....
[...]
References
33,771 citations
19,056 citations
13,033 citations
"Cogito componentiter ergo sum" refers background in this paper
...well-understood task when based on labelled examples [7]....
[...]
12,443 citations
"Cogito componentiter ergo sum" refers background in this paper
...Deerwester and co-workers developed the concept of latent semantics based on principal component analysis of the term-document matrix [15]....
[...]
8,333 citations
Related Papers (1)
Frequently Asked Questions (11)
Q2. What is the metric for term frequency?
A metric can be based on the simple Euclidean distance if document vectors are properly normalized, otherwise angular distance may be useful.
Q3. How do the authors filter out the inevitable noise?
By projecting the term-frequency vectors on a relatively low dimensional subspace, say determined by the maximal amount of variance one would be able to filter out the inevitable ‘noise’.
Q4. What is the meaning of the term-document matrix?
The vector space representation can be used for classification and retrieval by noting that similar documents are somehow expected to be ‘close’ in the vector space.
Q5. What is the purpose of the research programme?
The authors are currently pursuing a research programme, trying to understand the limitations of the ecological hypothesis for higher level cognitive processes, such as grouping abstract objects, navigating social networks, understanding multi-speaker environments, and understanding the representational differences between self and environment.
Q6. What is the simplest way to make the visualization independent of pitch?
To make the visualization relatively independent of ‘pitch’, the authors use the so-called mel-cepstral representation (MFCC, K = 13 coefficients pr. frame).
Q7. What is the fundamental observation behind the latent semantic indexing approach?
The fundamental observation behind the latent semantic indexing (LSI) approach is that similar documents are using similar vocabularies, hence, the vectors of a given topic could appear as produced by a stochastic process with highly correlated term-entries.
Q8. What are the eigenvectors of the matrix?
The eigenvectors of this matrix are called ‘eigencasts’ and represent characteristic communities of actors that tend to co-appear in movies.
Q9. What is the definition of a cognitive component analysis?
The authors thus define cognitive component analysis (COCA) as unsupervised grouping of data such that the ensuing group structure is well-aligned with that resulting from human cognitive activity [8].
Q10. What is the significance of the label structure in the real world?
It is a fascinating finding in many real world data sets that the label structure discovered by unsupervised learning closely coincides with labels obtained by letting a human or a group of humans perform classification, labels derived from human cognition.
Q11. How can the authors estimate consensus musical genre?
It is now feasible to estimate consensus musical genre by supervised learning from rather short music segments, say 5-10 seconds, see e.g., [18], thus enabling computerized handling of music request at a high cognitive complexity level.