search

NTMC-Community / MatchZoo

Facilitating the design, comparison and sharing of deep text matching models.
Apache License 2.0
3.83k stars 899 forks source link

dssm's ndcg , map metrics are less than the matchzoo tutorials #731

Closed Nicksphere closed 5 years ago

Nicksphere commented 5 years ago

I run the dssm algorithm with matchzoo 2.0，found that the result of dssm's ndcg, map metrics are less than matchzoo's tutorials obviously. Is there a problem ？

dataset : wiki environment: python3.6

my result: Validation: normalized_discounted_cumulative_gain@3(0.0): **0.18537827976306148** - normalized_discounted_cumulative_gain@5(0.0): **0.20825197021237893** - precision@1(0.0): 0.12164296998420221 - precision@3(0.0): 0.09268035808320169 - precision@4(0.0): 0.07780410742496051 - mean_reciprocal_rank(0.0): 0.19736153130702894 - mean_average_precision(0.0): **0.19227106490139664**

matchzoo tutorails result: Validation: normalized_discounted_cumulative_gain@3(0.0): 0.45663442312293695 - normalized_discounted_cumulative_gain@5(0.0): 0.5363645153841258 - mean_average_precision(0.0): 0.4956098197015037

uduse commented 5 years ago

You offer no clue, so I have no clue.

uduse commented 5 years ago

I hope things are working well for you now. I’ll go ahead and close this issue, but I’m happy to continue further discussion whenever needed.