This directory contains results updated in real time from an experiment about the partly supervised learning of representations and concepts.

The code is available on https://fleuret.org/git/culture and a summary of the whole experiment is available as a thread on X.com.

plots.png

performance graphs

w_quizzes/

examples of quizzes from the reference hand-designed tasks.

c_quizzes/

quizzes that have been created by the generative models.

prediction/

prediction on quizzes. Green and red frames indicate what has been predicted and if it is correct.

generation/

generation of quizzes.

other/

other files

20241026-01:44:37 test_loss 237 model 0 0.025384864807128905
20241026-01:44:37 quiz_set nb_w_quizzes 2500 nb_c_quizzes 0
20241026-01:44:39 test_accuracy 237 model 0 nb_correct 2070 / 2500 (82.80%)
20241026-01:44:39 quiz_set nb_w_quizzes 150 nb_c_quizzes 0
20241026-01:44:45 wrote reasoning.pth
20241026-01:44:46 quiz_set nb_w_quizzes 50000 nb_c_quizzes 0
20241026-01:47:19 train_loss 238 model 0 0.03170183181762695
20241026-01:47:19 quiz_set nb_w_quizzes 2500 nb_c_quizzes 0
20241026-01:47:21 test_loss 238 model 0 0.023565826416015626
20241026-01:47:21 quiz_set nb_w_quizzes 2500 nb_c_quizzes 0
20241026-01:47:23 test_accuracy 238 model 0 nb_correct 2064 / 2500 (82.56%)
20241026-01:47:23 quiz_set nb_w_quizzes 150 nb_c_quizzes 0
20241026-01:47:30 wrote reasoning.pth
20241026-01:47:30 quiz_set nb_w_quizzes 50000 nb_c_quizzes 0
20241026-01:50:02 train_loss 239 model 0 0.03240221405029297
20241026-01:50:02 quiz_set nb_w_quizzes 2500 nb_c_quizzes 0
20241026-01:50:04 test_loss 239 model 0 0.023733367919921877
20241026-01:50:04 quiz_set nb_w_quizzes 2500 nb_c_quizzes 0
20241026-01:50:07 test_accuracy 239 model 0 nb_correct 2049 / 2500 (81.96%)
20241026-01:50:07 quiz_set nb_w_quizzes 150 nb_c_quizzes 0
20241026-01:50:13 wrote reasoning.pth
20241026-01:50:13 quiz_set nb_w_quizzes 50000 nb_c_quizzes 0
20241026-01:52:47 train_loss 240 model 0 0.03212994384765625
20241026-01:52:47 quiz_set nb_w_quizzes 2500 nb_c_quizzes 0
20241026-01:52:49 test_loss 240 model 0 0.025578842163085938
20241026-01:52:49 quiz_set nb_w_quizzes 2500 nb_c_quizzes 0
20241026-01:52:51 test_accuracy 240 model 0 nb_correct 2047 / 2500 (81.88%)
20241026-01:52:51 quiz_set nb_w_quizzes 150 nb_c_quizzes 0
20241026-01:52:58 wrote reasoning.pth
20241026-01:52:58 quiz_set nb_w_quizzes 50000 nb_c_quizzes 0