-
Notifications
You must be signed in to change notification settings - Fork 201
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Data structures for approximate queries. #88
Conversation
Test FAILed. |
Test FAILed. |
Test PASSed. |
Test PASSed. |
Test FAILed. |
* Generic interface for arbitrary types. * Hash manager for sharing hash functions between sketches. * No concurrency guarantees yet.
* Templated counter type for potentially better configurable space savings.
* Can use thread-local median buffers for estimate().
* Less support for arbitrary T. * But faster hashing with std::hash.
* Size of sketch and accuracy of estimates are correlated. * Verified that estimates are not guaranteed to be non-negative and that this is correct behavior.
Test PASSed. |
* Using PQ to track heavy hitters gives precise results * Using PQ approximation (hash map) gives less predictable results
Test PASSed. |
Test PASSed. |
* Measure errors for elements with frequencies exceeding alpha threshold (count sketch). This shows a much more reasonable trend in the errors (parameters respected). Doesn't make sense to measure median over all errors since this includes infrequent elements. * Zipf handled well by universal sketch. * Need to rewrite tests / remove heavy computation from them and put somewhere else (only test correctness).
Test FAILed. |
* Increasing the number of layers only helps until a point, and then query accuracy decreases(although seems to still be within guaranteed bounds under current tests). * Minor changes. * Removed class templating. * Commented out univmon dependency.
Test FAILed. |
Test FAILed. |
Test FAILed. |
Test FAILed. |
Test FAILed. |
Test FAILed. |
Test FAILed. |
Test PASSed. |
Test PASSed. |
Test PASSed. |
Test PASSed. |
Test FAILed. |
Test FAILed. |
Test PASSed. |
Closing, will create new PR. |
PR for #86.