Roadmap » History » Version 17

Jeremy Gow, 2013-01-30 02:43 PM

1 1 Jeremy Gow
h1. Roadmap
2 1 Jeremy Gow
3 16 Jeremy Gow
The development code has been made compatible with sbcl 1.1.  We intend to release a private beta, once an appropriate license 
4 16 Jeremy Gow
has been chosen, then work toward a public beta and/or a web service.
5 1 Jeremy Gow
6 9 Jeremy Gow
h2. Short-term
7 1 Jeremy Gow
8 16 Jeremy Gow
Fix built-in examples (Conklin 95 etc.): data doesn't contain all the necessary basic viewpoints.
9 1 Jeremy Gow
10 16 Jeremy Gow
New viewpoint: metrical accent.
11 16 Jeremy Gow
12 1 Jeremy Gow
Viewpoint selection:
13 5 Jeremy Gow
* Adding viewpoint weights to output.
14 13 Jeremy Gow
* Print trace information about VP sets being tested + mean IC values; record this to log file.
15 16 Jeremy Gow
* Optionally specify: min-links
16 1 Jeremy Gow
* More flexible way for user to specify constraints on viewpoint search:
17 1 Jeremy Gow
** Define labelled viewpoint classes
18 1 Jeremy Gow
** Pairs/triples of labels/wildcards specify acceptable combinations
19 1 Jeremy Gow
** User provides whitelist or blacklist spec
20 1 Jeremy Gow
21 1 Jeremy Gow
Efficiency:
22 6 Jeremy Gow
* Check/extend caching of models etc.
23 1 Jeremy Gow
* Use sampling to estimate mean IC during VP selection.
24 1 Jeremy Gow
25 1 Jeremy Gow
Some basic benchmarks to ensure stability of future development versions.
26 14 Jeremy Gow
27 11 Jeremy Gow
28 11 Jeremy Gow
h2. Mid-term goals
29 10 Jeremy Gow
30 17 Jeremy Gow
Review derived viewpoints: many depend on MIDI pitch representation, incompatible with cents.
31 17 Jeremy Gow
32 10 Jeremy Gow
Make system more data agnostic:
33 10 Jeremy Gow
* Remove dependancy on amuse.
34 10 Jeremy Gow
* Separation of music viewpoints from model.
35 10 Jeremy Gow
* A straightforward interface for specifying viewpoints, including viewpoint schemas (e.g. interval, interval size)
36 1 Jeremy Gow
37 1 Jeremy Gow
Viewpoint selection:
38 4 Jeremy Gow
* Optimise based on match with existing IC values.
39 5 Jeremy Gow
40 5 Jeremy Gow
Allow user to specify structure of model.
41 4 Jeremy Gow
* Determine order in which distributions are combined.
42 5 Jeremy Gow
* Specify weights for particular combinations, e.g. weighted viewpoints, or weighted memory stores.
43 5 Jeremy Gow
* Multiple memory stores.
44 5 Jeremy Gow
* Specify alternative context strategies.
45 5 Jeremy Gow
* Provide some prepackaged models, e.g. the current model structure.
46 5 Jeremy Gow
47 5 Jeremy Gow
Allow models to use predictive information (PI), expected PI and PI rate (as analogs to IC, entropy and entropy rate respectively).
48 1 Jeremy Gow
49 15 Jeremy Gow
Predict more than one basic viewpoint, and provide recommended viewpoints for each one (not just cpitch and bioi). 
50 15 Jeremy Gow
51 15 Jeremy Gow
Predict over more than one dataset.
52 1 Jeremy Gow
53 1 Jeremy Gow
h2. Long-term
54 1 Jeremy Gow
55 14 Jeremy Gow
Hierarchical structure: chunk common patterns into symbols.
56 7 Jeremy Gow
Parallel implementation.
57 1 Jeremy Gow
58 14 Jeremy Gow
h2. Minor problems to fix
59 9 Jeremy Gow
60 14 Jeremy Gow
* It is possible to import empty datasets, which cause an error when described.
61 12 Jeremy Gow
* Create root directory structure if not present.
62 9 Jeremy Gow
* Show all viewpoints in describe-dataset.
63 9 Jeremy Gow
* Zero barlength (e.g. in Makam data) causes divide by zero error.
64 9 Jeremy Gow
* Update kern import to handle new pitch viewpoints.