Add bias term by default

[kmaehashi]

As reported in the mailing list and Casual Talk event, some users want this.
We should discuss we should support it or not.

[hido]

It is just a problem of selection. Implementation is easy and straightforward.

Merit of supporting bias term:

• Users can use bias term without care nor adding constant feature value
• System can reduce the networking cost by omitting constant feature value

Disadvantage of supporting bias term:

• Users cannot use linear models without bias term
  • The corresponding weight for the bias term should converge to zero if the true model does not have such constant term
  • But it may affect the accuracy especially in the early stage of training
• Another bias term mistakenly-added by user may cause the multicollinearity problem and degrade the accuracy

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[stale]

Closing this issue after a prolonged period of inactivity. Please feel free to create a new issue or re-open it if you need.