Let's discuss how are you using igel and which updates/features would you rather see in the future about igel

[dinhanhx]

@Anenizer

I think most of the work would be the possibility to use it as a Library, or what do you think?

According to igel README.rst Quick Start, the last bullet,

I think you can treat it as library

[dinhanhx]

Feature: there is an option to export diagrams from sklearn visualizations

[nidhaloff]

@dinhanhx I see what you mean. Actually, I opened this issue to receive people's feedback. I wanted a user to request a feature so that I can take it into consideration. I know some models like IsotonicRegression have not been added but are there users out there who will use the IsotonicRegression model if it was implemented? this is the real question here.

Therefore, I opened this issue not to received a developer's feedback but rather a user's feedback. It's not about brainstorming but rather actual users should say what they like/don't like about igel. For example, if a user write down that IsotonicRegression is very important for him/her as he/she is using it in daily life then I would happily start working on it. However, I don't think that is realistic. The implemented models are more than enough to make powerful machine learning stuff.

From my experience, people who are using such models (like IsotonicRegression and the like) are more into research and want to go deep down to the details so they would probably use sklearn instead of igel since they don't need that abstraction.

However, you can work on this if you think it's important. It should be straightforward to add actually ;)

[KersyTom]

i think igel can accept other formatted data like a table from hive which contain features and target,we can just write down a table name and choose some columns,so then train a model conveniently

[dinhanhx]

There is much more models in sklearn, I think there should be an issue or a live website to keep track of implementing progress.

[nidhaloff]

@dinhanhx The visualization feature sounds good. What do you mean with your second comment?

[dinhanhx]

@nidhaloff For my second comment, I have perceived that igel has supported a lot of models the list. However, there are more models of sklearn which haven't been implemented in such as IsotonicRegression. Therefore, I think we should have a checklist/todo list (as a Github issue or Github project or a website) for models we are working on, we are going to do, etc.

[dinhanhx]

@nidhaloff I got your point now. I should wait for other people too.

[clausnizer-ondics]

Are there any plans in to interface igel in Jupyter to it be used in a Jupyter Notebook?
I think most of the work would be the possibility to use it as a Library, or what do you think?

[nidhaloff]

@KersyTom Thanks for the suggestion. Can you explain more about hive and how it handles data format. As far as I know, you can export the data in csv or excel format from hive, which igel also supports. Is this use case special? and if yes how exactly. It would be great if you provide more information about it

[KersyTom]

@nidhaloff In order to ensure customer privacy, sometimes downloading data is not allowed, or even viewing the data,but i know each column represent .My problem is How can I use sensitive data normally?

[nidhaloff]

@KersyTom I see. What do you mean by how can you use sensitive data normally? Igel does not store any data locally nor send information about your data to external sources. Your data will be used locally just by you. does that answer your question?

[nidhaloff]

discussion moved to #71