[REVIEW] Adding power_t param to SGD failing pytests

[divyegala]

SGDClassifier pytests were failing in CI for the specific test params, lrate=invscaling and loss=log. This was occurring due to floating point errors where values like 0.50xx would cause a misclassification. To resolve this, I simply increased the learning rate of the problem. For invscaling, lrate = eta0 / (pow(t, power_t), where eta0 and power_t are configurable parameters. Reducing power_t to 0.4 from its default of 0.5 was enough for the solution to converge for these pytests.

[GPUtester]

Please update the changelog in order to start CI tests.

View the gpuCI docs here.

[codecov-io]

Codecov Report

Merging #3012 into branch-0.17 will not change coverage.
The diff coverage is n/a.

@@             Coverage Diff              @@
##           branch-0.17    #3012   +/-   ##
============================================
  Coverage        58.87%   58.87%           
============================================
  Files              143      143           
  Lines             9004     9004           
============================================
  Hits              5301     5301           
  Misses            3703     3703           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 95a7305...b473878. Read the comment docs.