RetroC temporal classification challenge

Guess the publication year of a Polish text.

submitter when description dev-0/RMSE test-A/RMSE
[anonymised] 2016-06-11 14:29 Max entropy solution for 4 grams and 5 class 31.5 41.0
[anonymised] 2016-06-11 14:09 neural network, 4grams, 8 class 31.5 41.0
[anonymised] 2016-06-11 14:03 GLM 31.5 41.0
[anonymised] 2016-06-11 13:16 GLM solution for 4 grams and 8 class 31.5 41.0
R.J. 2016-05-29 14:08 VW -nn 6 on up to 4-grams and [5-7] tokens + wiki years 17.1 24.9
R.J. 2016-05-29 10:35 Wiki years - min year from article, with [184-2013] 81.8 82.9
R.J. 2016-05-29 08:21 Wiki years 103.1 N/A
R.J. 2016-05-29 07:48 Wiki years 103.1 N/A
R.J. 2016-05-25 13:13 VW -nn 6 on up to 5-grams and [6-8] tokens 17.9 26.4
R.J. 2016-05-25 11:40 Current best solution with changed learning rate 17.3 25.5
[anonymised] 2016-05-13 13:21 Max entropy solution for 4 grams and 8 class 26.6 35.8
[anonymised] 2016-05-13 13:16 Solution for 4grams and 8 class 24.7 36.8
[anonymised] 2016-05-07 21:29 Solution for 4grams and 10 class 24.7 37.4
[anonymised] 2016-05-07 21:25 Solution for 4grams and 20 class 25.1 38.4
[anonymised] 2016-04-30 18:20 Solution for 4grams and 40 class 26.7 40.3
[anonymised] 2016-04-29 20:21 Solution for 4grams and 67 class 27.3 41.6
[anonymised] 2016-04-29 07:27 Solution for 4grams and 100 class 28.4 41.8
[anonymised] 2016-04-22 11:32 Solution for 4grams 30.2 43.2
[anonymised] 2016-04-20 12:19 Solution 30.2 44.6
[anonymised] 2016-01-07 07:56 Po najczęściej występujących słowach N/A N/A
p/tlen 2015-12-19 11:07 VW classifier (10-year chronon) 24.5 34.4
[anonymised] 2015-12-16 23:38 first submission N/A 75.0
Marcin Junczys-Dowmunt 2015-12-16 20:40 another try with more bits 16.7 27.2
[anonymised] 2015-12-16 20:13 + script 34.7 50.3
[anonymised] 2015-12-14 08:01 tfidf, min_df = 3 34.7 50.3
[anonymised] 2015-12-13 22:16 test with min_df = 3 41.5 60.1
Marcin Junczys-Dowmunt 2015-12-13 18:50 Simple 2-layer network with up to 4 chars n-grams 16.8 27.6
p/tlen 2015-12-13 14:31 VW -nn 6 on up to 4-grams and [5-7] tokens 17.2 24.8
p/tlen 2015-12-13 12:50 VW -nn 6 on up to 4-grams and [5-8] tokens 17.2 25.3
p/tlen 2015-12-13 12:49 VW -nn 6 on up to 4-grams and [5-8] tokens 45.9 46.5
[anonymised] 2015-12-13 11:52 Corrected better solution 31.9 44.4
[anonymised] 2015-12-13 11:43 Maybe better solution N/A 44.4
Marcin Junczys-Dowmunt 2015-12-12 22:02 The same as last, best epoch 16.3 24.9
[anonymised] 2015-12-12 15:56 test samego dev-0 dla okrojonego treningu 61.5 N/A
p/tlen 2015-12-12 15:12 up to 4-grams (VW with -nn 6 found with vw-hypersearch) 17.2 25.5
p/tlen 2015-12-12 15:10 up to 4-grams (VW with -nn 6 found with vw-hypersearch) 45.9 46.5
Marcin Junczys-Dowmunt 2015-12-11 19:49 Same as lat, word2vec pretraining 16.4 25.3
Marcin Junczys-Dowmunt 2015-12-11 19:37 200 hidden units in GRU, more epochs, full batches 17.5 27.3
[anonymised] 2015-12-11 16:32 First solution 30.2 44.6
p/tlen 2015-12-10 21:15 up to 4-grams (VW with -nn 10) 17.7 26.0
Marcin Junczys-Dowmunt 2015-12-09 08:46 900 words, more units, RMSprop 18.5 27.5
p/tlen 2015-12-08 20:18 signi tempori (with poor man's stemming by taking the first 7 letters) 48.1 50.2
Marcin Junczys-Dowmunt 2015-12-07 21:26 Better regression layer, 10 year bins weighted average 19.6 30.0
Marcin Junczys-Dowmunt 2015-12-07 14:58 just one epoch, to avoid overfitting 23.3 33.8
Marcin Junczys-Dowmunt 2015-12-07 14:50 Much simpler model, less hidden units, l2=1e-5 20.8 31.7
Marcin Junczys-Dowmunt 2015-12-07 09:20 now with better regularization 20.9 31.6
Marcin Junczys-Dowmunt 2015-12-04 22:47 Neural Network, first try 22.0 30.2
p/tlen 2015-11-25 22:29 up to 4-grams 19.6 28.4
p/tlen 2015-11-25 22:27 up to 4-grams 45.9 46.5
p/tlen 2015-11-25 20:47 5-grams 22.0 33.5
p/tlen 2015-11-25 20:05 birth years 57.4 56.7
p/tlen 2015-11-24 22:23 hand-crafted rules 46.6 50.9
p/tlen 2015-11-24 21:40 by known words 57.9 56.5
p/tlen 2015-11-24 20:24 by year references 45.9 46.5
p/tlen 2015-11-24 19:59 null model with a half year 57.9 57.7
p/tlen 2015-11-24 19:58 null model 57.9 57.7
p/tlen 2015-11-14 16:30 test 81.1 57.7