qnstrn was invoked with the following argument values (some of which may be defaults): ftr1_file=/u/drspeech/data/aurora/pfile/plp12/a2-train-rand.pf \ ftr1_format=pfile \ ftr1_width=0 \ ftr2_file= \ ftr2_format=pfile \ ftr2_width=0 \ unary_file= \ hardtarget_file=/u/drspeech/data/aurora/label/a2-plp+msg-2/a2-train-rand.ilab \ hardtarget_format=ilab \ softtarget_file= \ softtarget_format=pfile \ softtarget_width=0 \ ftr1_norm_file=/u/drspeech/data/aurora/pfile/plp12/train-rand.norms \ ftr2_norm_file= \ ftr1_ftr_start=0 \ ftr2_ftr_start=0 \ ftr1_ftr_count=39 \ ftr2_ftr_count=0 \ window_extent=9 \ ftr1_window_offset=0 \ ftr2_window_offset=4 \ unary_window_offset=3 \ hardtarget_window_offset=4 \ softtarget_window_offset=0 \ ftr1_window_len=9 \ ftr2_window_len=0 \ ftr1_delta_order=2 \ ftr1_delta_win=9 \ ftr1_norm_utts=true \ ftr2_delta_order=0 \ ftr2_delta_win=9 \ ftr2_norm_utts=false \ train_cache_frames=10000 \ train_cache_seed=0 \ train_sent_start=800 \ train_sent_count=7640 \ cv_sent_start=0 \ cv_sent_count=800 \ init_random_bias_min=-4.1 \ init_random_bias_max=-3.9 \ init_random_weight_min=-0.1 \ init_random_weight_max=0.1 \ init_random_seed=0 \ init_weight_file= \ log_weight_file=swbt-plp12ddN-ta2multi-aa2-plp+msg-2+351i+480h+56o-%e.wts \ out_weight_file=boot-plp12ddN-ta2multi-aa2-plp+msg-2+351i+480h+56o.wts \ learnrate_schedule=newbob \ learnrate_vals=0.008 \ learnrate_epochs=12 \ learnrate_scale=0.5 \ unary_size=0 \ mlp3_input_size=351 \ mlp3_hidden_size=480 \ mlp3_output_size=56 \ mlp3_output_type=softmax \ mlp3_fx=true \ mlp3_weight_bits=32 \ mlp3_in2hid_exp=2 \ mlp3_hid2out_exp=2 \ mlp3_bunch_size=16 \ slaves=2 \ cpu=spert \ log_file=swbt-plp12ddN-ta2multi-aa2-plp+msg-2+480h.log \ verbose=false \ debug=0 \ Program version: qnstrn v0_97a. Host: ravioli. Program start: Mon May 22 12:54:32 2000. Training run start: Mon May 22 12:54:37 2000. ** ** ** ** ** ** ** ** ** ** ** ** ** ** Pre-run cross validation started: Mon May 22 12:54:37 2000. CV speed: 162.89 MCPS, 831.5 presentations/sec. CV accuracy: 712 right out of 134061, 0.53% correct. ** ** ** ** ** ** ** ** ** ** ** ** ** ** New epoch: epoch 1, learn rate 0.008000. Epoch started: Mon May 22 12:57:19 2000. Train speed: 109.63 MCUPS, 559.6 presentations/sec. Train accuracy: 995192 right out of 1267377, 78.52% correct. CV speed: 160.94 MCPS, 821.6 presentations/sec. CV accuracy: 100660 right out of 134061, 75.09% correct. Weight log: Saving weights to `swbt-plp12ddN-ta2multi-aa2-plp+msg-2+351i+480h+56o-1.wts'. ** ** ** ** ** ** ** ** ** ** ** ** ** ** New epoch: epoch 2, learn rate 0.008000. Epoch started: Mon May 22 13:37:50 2000. Train speed: 109.54 MCUPS, 559.2 presentations/sec. Train accuracy: 1042160 right out of 1267377, 82.23% correct. CV speed: 160.99 MCPS, 821.8 presentations/sec. CV accuracy: 102309 right out of 134061, 76.32% correct. Weight log: Saving weights to `swbt-plp12ddN-ta2multi-aa2-plp+msg-2+351i+480h+56o-2.wts'. ** ** ** ** ** ** ** ** ** ** ** ** ** ** New epoch: epoch 3, learn rate 0.008000. Epoch started: Mon May 22 14:18:24 2000. Train speed: 109.51 MCUPS, 559.0 presentations/sec. Train accuracy: 1052872 right out of 1267377, 83.07% correct. CV speed: 160.97 MCPS, 821.7 presentations/sec. CV accuracy: 103049 right out of 134061, 76.87% correct. Weight log: Saving weights to `swbt-plp12ddN-ta2multi-aa2-plp+msg-2+351i+480h+56o-3.wts'. ** ** ** ** ** ** ** ** ** ** ** ** ** ** New epoch: epoch 4, learn rate 0.008000. Epoch started: Mon May 22 14:58:58 2000. Train speed: 109.67 MCUPS, 559.9 presentations/sec. Train accuracy: 1058468 right out of 1267377, 83.52% correct. CV speed: 161.11 MCPS, 822.4 presentations/sec. CV accuracy: 103089 right out of 134061, 76.90% correct. Weight log: Saving weights to `swbt-plp12ddN-ta2multi-aa2-plp+msg-2+351i+480h+56o-4.wts'. ** ** ** ** ** ** ** ** ** ** ** ** ** ** New epoch: epoch 5, learn rate 0.004000. Epoch started: Mon May 22 15:39:28 2000. Train speed: 109.57 MCUPS, 559.3 presentations/sec. Train accuracy: 1063144 right out of 1267377, 83.89% correct. CV speed: 161.16 MCPS, 822.7 presentations/sec. CV accuracy: 106164 right out of 134061, 79.19% correct. Weight log: Saving weights to `swbt-plp12ddN-ta2multi-aa2-plp+msg-2+351i+480h+56o-5.wts'. ** ** ** ** ** ** ** ** ** ** ** ** ** ** New epoch: epoch 6, learn rate 0.002000. Epoch started: Mon May 22 16:20:01 2000. Train speed: 109.43 MCUPS, 558.6 presentations/sec. Train accuracy: 1067775 right out of 1267377, 84.25% correct. CV speed: 160.93 MCPS, 821.5 presentations/sec. CV accuracy: 107479 right out of 134061, 80.17% correct. Weight log: Saving weights to `swbt-plp12ddN-ta2multi-aa2-plp+msg-2+351i+480h+56o-6.wts'. ** ** ** ** ** ** ** ** ** ** ** ** ** ** New epoch: epoch 7, learn rate 0.001000. Epoch started: Mon May 22 17:00:37 2000. Train speed: 109.72 MCUPS, 560.1 presentations/sec. Train accuracy: 1071052 right out of 1267377, 84.51% correct. CV speed: 160.81 MCPS, 820.9 presentations/sec. CV accuracy: 108418 right out of 134061, 80.87% correct. Weight log: Saving weights to `swbt-plp12ddN-ta2multi-aa2-plp+msg-2+351i+480h+56o-7.wts'. ** ** ** ** ** ** ** ** ** ** ** ** ** ** New epoch: epoch 8, learn rate 0.000500. Epoch started: Mon May 22 17:41:07 2000. Train speed: 109.55 MCUPS, 559.2 presentations/sec. Train accuracy: 1073079 right out of 1267377, 84.67% correct. CV speed: 161.09 MCPS, 822.3 presentations/sec. CV accuracy: 108748 right out of 134061, 81.12% correct. Weight log: Saving weights to `swbt-plp12ddN-ta2multi-aa2-plp+msg-2+351i+480h+56o-8.wts'. ** ** ** ** ** ** ** ** ** ** ** ** ** ** Best CV accuracy: 81.12% correct in epoch 8. Best train accuracy: 84.67% correct in epoch 8. Training run stop: Mon May 22 18:21:40 2000. Training time: 19622.76 secs (5 hours, 27 mins, 2 secs). ** ** ** ** ** ** ** ** ** ** ** ** ** ** Weights written to 'boot-plp12ddN-ta2multi-aa2-plp+msg-2+351i+480h+56o.wts'. Program stop: Mon May 22 18:21:44 2000