Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- Intel Community
- Software
- Software Development SDKs and Libraries
- Intel® oneAPI Data Analytics Library
- Regression and time series forecasting

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Mute
- Printer Friendly Page

janez-makovsek

Beginner

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

07-26-2017
12:41 AM

76 Views

Regression and time series forecasting

Hi,

Except for linear regression and ridge-regression, are there are algorithms (with examples) in DAAL which could be used for regression of an arbitrary function without an explicit model? What about time-series forecasting?

Thanks!

Atmapuri

Link Copied

3 Replies

Andrey_N_Intel

Employee

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

07-26-2017
02:40 AM

76 Views

Hi Atmapuri,

The present version of the library supports linear and ridge regression only. Can you provide additional details related to your first question - specific algorithm(s) you keep in mind, typical input problem sizes, precision used to represent the input data, the need in online and/or distributed mode of the computations?

For the time series topic - indeed, we analyze support of the functionality in the library, and your inputs similar to the ones above on the specific algorithms/use cases would be helpful as well.

Thanks,

Andrey

janez-makovsek

Beginner

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

07-27-2017
03:49 AM

76 Views

Dear Andrey,

Within AI and machine learning the algorithms are rarely very generic. They mostly apply to some specific type of problem and work well with that. It would be great, if you could include at least one real world example per algorithm rather than random generated numbers. And add short description of the problem (types) being solved.

Especially in the case of Neural Networks it would help a lot to have a prebuilt layer setups, which only need to be modified for feature count and type for example, but could be used directly. Depending on the area of application there could be hundreds of different configurations. Having ready-to-use-configs with real life data would make a big difference. Examples for NN:

Binary Classification

Multi-class classifications with ordinal, categorical and real valued inputs from tables.

Handling of missing values with NNs

Regression example (real valued output with all three possible input feature types an real valued output )

At least three photo classification examples.

Time series prediction.

...

Thanks!

Atmapuri

Andrey_N_Intel

Employee

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

07-27-2017
04:58 AM

76 Views

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

For more complete information about compiler optimizations, see our Optimization Notice.