Benchmark recommenders with fit! optimization and refactoring
#50
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Latest benchmark result: ```julia julia> using BenchmarkTools julia> using Recommendation julia> data = load_movielens_100k() julia> recommender = MostPopular(data) julia> @Btime fit!(recommender) 32.667 ms (2976 allocations: 59.84 KiB) ``` whereas the result originally was: ```julia julia> @Btime fit!(recommender) 58.305 ms (2322995 allocations: 47.77 MiB) ```
Codecov Report
@@ Coverage Diff @@
## master #50 +/- ##
==========================================
- Coverage 79.29% 78.63% -0.67%
==========================================
Files 23 23
Lines 739 716 -23
==========================================
- Hits 586 563 -23
Misses 153 153
Continue to review full report at Codecov.
|
`fit!` performance (before vs. after): CoOccurrence [need improvement] 108.658 ms (4658986 allocations: 96.15 MiB) 181.440 ms (1589204 allocations: 36.40 MiB) ItemMean 98.345 ms (5500291 allocations: 96.25 MiB) 50.346 ms (3180166 allocations: 48.54 MiB) UserMean 129.993 ms (5870962 allocations: 101.84 MiB) 88.428 ms (5488286 allocations: 83.76 MiB) ThresholdPercentage 124.990 ms (5500295 allocations: 168.93 MiB) 117.950 ms (3176480 allocations: 121.13 MiB)
``` using Recommendation using BenchmarkTools data = load_movielens_100k(); recommender = CoOccurrence(data, 1); fit!(recommender); ``` Before 176.925 ms (1589204 allocations: 36.40 MiB) After 51.136 ms (1589140 allocations: 24.48 MiB)
``` data = DataAccessor(reshape(collect(1:1000), 20, 50)) recommender = UserKNN(data) fit!(recommender) ``` Before 796.582 μs (8632 allocations: 1.55 MiB) After 565.204 μs (6537 allocations: 760.14 KiB) No significant change in `predict` for now, but add a line to avoid `k` exceeds total user size.
``` data = load_movielens_100k() recommender = ItemKNN(data) @Btime fit!(recommender, adjusted_cosine=true) ``` Before (process doesn't finish in minutes) After 17.048 s (33367232 allocations: 20.75 GiB) No significant change in `predict` for now, but add a line to avoid `k` exceeds total item size.
e.g., `n_user` -> `n_users`, `n_factor` -> `n_factors`
|
Quick benchmark results on a local laptop: Baselines: julia> using Recommendation
julia> using BenchmarkTools
julia> data = load_movielens_100k();
julia> recommender = UserMean(data);
julia> @btime fit!(recommender);
88.811 ms (5489223 allocations: 83.77 MiB)
julia> recommender = ItemMean(data);
julia> @btime fit!(recommender);
51.546 ms (3180166 allocations: 48.54 MiB)
julia> recommender = MostPopular(data);
julia> @btime fit!(recommender);
35.052 ms (2976 allocations: 59.84 KiB)
julia> recommender = ThresholdPercentage(data, 3.0);
julia> @btime fit!(recommender);
110.403 ms (3176482 allocations: 121.13 MiB)
julia> recommender = CoOccurrence(data, 1);
julia> @btime fit!(recommender);
51.660 ms (1589140 allocations: 24.48 MiB)Model-based: julia> recommender = UserKNN(data, 20, true);
julia> @btime fit!(recommender);
9.414 s (125878781 allocations: 2.91 GiB)
julia> recommender = ItemKNN(data);
julia> @btime fit!(recommender, adjusted_cosine=true);
16.051 s (33367232 allocations: 20.75 GiB)
julia> recommender = SVD(data);
julia> @btime fit!(recommender);
343.518 ms (1638163 allocations: 76.85 MiB)
julia> recommender = MatrixFactorization(data);
julia> @btime fit!(recommender, shuffled=false);
28.784 s (329080643 allocations: 32.32 GiB)
julia> recommender = FactorizationMachines(data);
julia> @btime fit!(recommender, shuffled=false);
138.807 s (1993371437 allocations: 215.60 GiB) |
takuti
changed the title
fit!fit! optimization and refactoring
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Relates to #26