Here is a summary of the responses:
7 C++ core with concrete API
- encoder extension points
- no library dependencies
6 Real temporal pooling
5 Hierarchies of CLA regions
4 Compelling sample application
3 Language-independent serialization format
3 CMake support
3 Multiple variable prediction with arbitrary time intervals
2 Sensorimotor model
1 Binary distros for common platforms
1 Pip install nupic
1 Comparisons with other ML techniques
1 Full descriptions of algorithms
1 Regression framework to measure algorithmic progress
1 Vision focus
1 Optimize TP for GPU and multicore
1 Repository restructure
---------
Matt Taylor
OS Community Flag-Bearer
Numenta
On Tue, Jan 7, 2014 at 3:57 AM, David Ragazzi <davidraga...@gmail.com> wrote:
> +1 Organize the repository structure:
> I think we should have intuitive and straight names to Nupic subprojects.
> For example, instead of the core be in NTA/algoritms, I think we should have
> a exclusive folder only with this project (CLA for example), OPF (frameworks
> folder) also shouldn't be as subfolder as it is a external project that
> consumes the CLA. Also is interesting organize the files in Source, Release,
> Docs, and Build_System folders (although some progress already was made),
> because this structure is easy to understand to newbies and also separate
> the components according to specific public.
>
> An suggestion of folders structure:
> 1. Nupic
> 1.1. Source: source files of all subprojects
> 1.1.1. CLA: project folder of the stand-alone CLA algorithm
> 1.1.2. OPF: project folder of the prediction framework
> 1.1.3. Encoders: folder containing encoders (text, numbers, etc) with
> inherited classes to manipulate HTM input and abstract the process.
> 1.1.4. DataSources: folder containing data sources (vector, database,
> text files, etc) to abstract the data access.
> 1.2. Doc: documentation folder including pdfs, videos, etc
> 1.2.1. User: documentation for those that only want test and integrate
> Nupic with their projects.
> 1.2.2. Developer: documentation for those that want put hands on. It
> could include intro about project, convention code, development process,
> etc.
> 1.3. Relase: temporary folder containing all binary files generated by
> building process. This also could allow that Nupic be downloaded in 2
> categories: Release (for attending users) or Source+Release (for attending
> developers)
> 1.4. Build_System: temporary folder containing the build system generated
> by Cmake.
>
> Well, this is only a rough draft.. but it´s my 1 cent.. :-)
>
> David
>
>
> On 6 January 2014 22:53, Matthew Taylor <m...@numenta.org> wrote:
>>
>> Hello NuPIC community,
>>
>> 2014 is here, and I'd like to start this year off by getting some
>> ideas and feedback from you. Just one simple question:
>>
>> What is the most important work that needs to be done for NuPIC this
>> year?
>>
>> We've already had a lot of discussion about this in several threads on
>> the mailing list over the past 6 months, but I don't want to miss
>> anything. Tell me what YOU think is important, either directly to
>> m...@numenta.org, but preferably to the mailing list so we can discuss
>> it.
>>
>> Is it a well-defined core API? Is it a C++ only implementation of the
>> core algorithms and encoders? A pip-installable NuPIC? More complete
>> documentation and example code? More hackathons and videos?
>> Comparisons with other ML algorithm implementations?
>>
>> I'm looking forward to hearing your opinions.
>>
>> Thanks,
>> ---------
>> Matt Taylor
>> OS Community Flag-Bearer
>> Numenta
>>
>> _______________________________________________
>> nupic mailing list
>> nupic@lists.numenta.org
>> http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
>
>
>
> _______________________________________________
> nupic mailing list
> nupic@lists.numenta.org
> http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
>
_______________________________________________
nupic mailing list
nupic@lists.numenta.org
http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
