Roadmap » History » Version 31

Marcus Pearce, 2014-06-19 06:29 PM

1 1 Jeremy Gow
h1. Roadmap
2 1 Jeremy Gow
3 23 Marcus Pearce
h2. Installation
4 20 Jeremy Gow
5 25 Marcus Pearce
* -Remove absolute pathname in connect-to-database. (mtp-admin/music-data.lisp)-
6 20 Jeremy Gow
* Create cache directories if they don't exist.
7 20 Jeremy Gow
8 23 Marcus Pearce
h2. Data import
9 1 Jeremy Gow
10 27 Marcus Pearce
* -It is possible to import empty datasets, which cause an error when described.-
11 31 Marcus Pearce
* running commands on an empty database produces an error (e.g., <code>mtp-admin:describe-database</code>)
12 20 Jeremy Gow
13 23 Marcus Pearce
h2. Viewpoints
14 20 Jeremy Gow
15 28 Marcus Pearce
* Polyphonic viewpoints for modelling harmonic movement
16 30 Marcus Pearce
* A straightforward language for specifying viewpoints, including viewpoint schemas (e.g. interval, interval size being target viewpoints), making system more data agnostic
17 23 Marcus Pearce
* Zero barlengths sometimes cause divide by zero errors
18 16 Jeremy Gow
19 23 Marcus Pearce
h2. Viewpoint selection
20 23 Marcus Pearce
21 23 Marcus Pearce
* Memory errors sometimes occur with large viewpoints sets (e.g., Cpitch with no basis specified)
22 23 Marcus Pearce
* When decimal places are restricted for comparison, earlier systems are preferred within a round, so use full precision to choose between ties
23 1 Jeremy Gow
* Print trace information about VP sets being tested + mean IC values; record this to log file.
24 16 Jeremy Gow
* Optionally specify: min-links
25 1 Jeremy Gow
* More flexible way for user to specify constraints on viewpoint search:
26 1 Jeremy Gow
** Define labelled viewpoint classes
27 1 Jeremy Gow
** Pairs/triples of labels/wildcards specify acceptable combinations
28 1 Jeremy Gow
** User provides whitelist or blacklist spec
29 28 Marcus Pearce
* Provide recommended viewpoints for more basic viewpoints than just cpitch, onset & bioi
30 17 Jeremy Gow
31 29 Marcus Pearce
h2. Modelling
32 21 Jeremy Gow
* Predict over more than one dataset.
33 29 Marcus Pearce
* Allow multiple memory stores (i.e., not restricted to just two: the LTM and STM).
34 23 Marcus Pearce
* Hierarchical structure: chunk common patterns into symbols using information content and entropy as indicators of grouping structure (Pearce et al., 2010, Perception)
35 29 Marcus Pearce
* Compute predictive information (PI), expected PI and PI rate as analogs to IC, entropy and entropy rate respectively (Abdallah & Plumbley, 2009).
36 29 Marcus Pearce
* Offer the selection of alternative context strategies (e.g., include future context)
37 29 Marcus Pearce
* Allow the user to fix the weights for particular combinations, e.g. weighted viewpoints, or weighted memory stores (LTM/STM etc.)
38 29 Marcus Pearce
* Allow the user to determine order of model combination (e.g., LTM-STM first, then viewpoints vs viewpoints first then LTM/STM)
39 23 Marcus Pearce
40 29 Marcus Pearce
h2. Longer-term goals
41 1 Jeremy Gow
42 29 Marcus Pearce
Interface
43 30 Marcus Pearce
* Implement as a web service to avoid installation woes.
44 29 Marcus Pearce
* Implement a 'Pace' viewpoint: measure of information rate (bits/sec), analogous to flow in speech production.
45 29 Marcus Pearce
46 23 Marcus Pearce
Efficiency:
47 15 Jeremy Gow
** Check/extend caching of models etc.
48 15 Jeremy Gow
** Use sampling to estimate mean IC during VP selection.
49 23 Marcus Pearce
** Optimise viewpoint selection based on match with existing IC values.
50 29 Marcus Pearce
51 29 Marcus Pearce
Testing
52 29 Marcus Pearce
** Include unit testing code.