[go: nahoru, domu]

tree: 378f863cbd4f240809a9d8f7eafe785a11b8c56d [path history] [tgz]
  1. proto/
  2. assist_ranker_service.h
  3. assist_ranker_service_impl.cc
  4. assist_ranker_service_impl.h
  5. base_predictor.cc
  6. base_predictor.h
  7. base_predictor_unittest.cc
  8. binary_classifier_predictor.cc
  9. binary_classifier_predictor.h
  10. binary_classifier_predictor_unittest.cc
  11. BUILD.gn
  12. classifier_predictor.cc
  13. classifier_predictor.h
  14. classifier_predictor_unittest.cc
  15. DEPS
  16. DIR_METADATA
  17. example_preprocessing.cc
  18. example_preprocessing.h
  19. example_preprocessing_unittest.cc
  20. fake_ranker_model_loader.cc
  21. fake_ranker_model_loader.h
  22. generic_logistic_regression_inference.cc
  23. generic_logistic_regression_inference.h
  24. generic_logistic_regression_inference_unittest.cc
  25. nn_classifier.cc
  26. nn_classifier.h
  27. nn_classifier_test_util.cc
  28. nn_classifier_test_util.h
  29. nn_classifier_unittest.cc
  30. OWNERS
  31. predictor_config.cc
  32. predictor_config.h
  33. predictor_config_definitions.cc
  34. predictor_config_definitions.h
  35. print_example_preprocessor_config.py
  36. quantized_nn_classifier.cc
  37. quantized_nn_classifier.h
  38. quantized_nn_classifier_unittest.cc
  39. ranker_example_util.cc
  40. ranker_example_util.h
  41. ranker_example_util_unittest.cc
  42. ranker_model.cc
  43. ranker_model.h
  44. ranker_model_loader.h
  45. ranker_model_loader_impl.cc
  46. ranker_model_loader_impl.h
  47. ranker_model_loader_impl_unittest.cc
  48. ranker_model_unittest.cc
  49. ranker_url_fetcher.cc
  50. ranker_url_fetcher.h
  51. README.md
components/assist_ranker/README.md

Assist Ranker

Introduction

Assist Ranker is design to make Chrome smarter by providing client-side machine learning (ML) inference in Chrome. It is designed to be a generic infrastructure that supports ML needs for all Chrome feature teams on all platforms.

Assist Ranker utilizes UKM logging to log per feature-label events. A ML model will be trained in the Cloud based on these logs; and then Assist Ranker will download and inference with the model.

It currently only supports Logistic Regression and Multilayer Neural Networks.

Assist Ranker was experiment on ContextualSearch; but it is not used in any production projects.

How to use it

Please contact the owners before you use it.