HW3: Source and results for both of the scenarios added
This commit is contained in:
@@ -0,0 +1,3 @@
|
||||
# Matlab auxiliary files
|
||||
*.asv
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,620 @@
|
||||
|
||||
Model 1: TSK0-2MF (gbellmf × 2 per input)
|
||||
=====================================
|
||||
Initial rules: 32
|
||||
|
||||
ANFIS info:
|
||||
Number of nodes: 92
|
||||
Number of linear parameters: 32
|
||||
Number of nonlinear parameters: 30
|
||||
Total number of parameters: 62
|
||||
Number of training data pairs: 902
|
||||
Number of checking data pairs: 301
|
||||
Number of fuzzy rules: 32
|
||||
|
||||
|
||||
Start training ANFIS ...
|
||||
|
||||
1 4.25643 4.51914
|
||||
2 4.22775 4.48492
|
||||
3 4.1994 4.45216
|
||||
4 4.17142 4.42081
|
||||
Step size increases to 0.011000 after epoch 5.
|
||||
5 4.14386 4.39082
|
||||
6 4.11676 4.36215
|
||||
7 4.08753 4.33208
|
||||
8 4.05896 4.3035
|
||||
Step size increases to 0.012100 after epoch 9.
|
||||
9 4.0311 4.27637
|
||||
10 4.004 4.25063
|
||||
11 3.97511 4.22389
|
||||
12 3.94724 4.19874
|
||||
Step size increases to 0.013310 after epoch 13.
|
||||
13 3.92044 4.17515
|
||||
14 3.89473 4.15306
|
||||
15 3.86777 4.13049
|
||||
16 3.84221 4.10967
|
||||
Step size increases to 0.014641 after epoch 17.
|
||||
17 3.81806 4.09058
|
||||
18 3.79532 4.0732
|
||||
19 3.77194 4.05604
|
||||
20 3.75022 4.04091
|
||||
Step size increases to 0.016105 after epoch 21.
|
||||
21 3.73013 4.0278
|
||||
22 3.7116 4.0167
|
||||
23 3.69293 4.00682
|
||||
24 3.67596 3.99934
|
||||
Step size increases to 0.017716 after epoch 25.
|
||||
25 3.66054 3.99425
|
||||
26 3.64656 3.99151
|
||||
27 3.63269 3.99118
|
||||
28 3.62021 3.99352
|
||||
Step size increases to 0.019487 after epoch 29.
|
||||
29 3.60898 3.99839
|
||||
30 3.59883 4.00557
|
||||
31 3.58878 4.01587
|
||||
32 3.57974 4.02819
|
||||
Step size increases to 0.021436 after epoch 33.
|
||||
33 3.57161 4.04196
|
||||
34 3.56429 4.05654
|
||||
35 3.55707 4.07282
|
||||
36 3.55066 4.08823
|
||||
Step size increases to 0.023579 after epoch 37.
|
||||
37 3.54498 4.10185
|
||||
38 3.53995 4.1129
|
||||
39 3.53508 4.12181
|
||||
40 3.53083 4.12685
|
||||
Step size increases to 0.025937 after epoch 41.
|
||||
41 3.5271 4.12846
|
||||
42 3.52381 4.12733
|
||||
43 3.5206 4.12391
|
||||
44 3.51772 4.11867
|
||||
Step size increases to 0.028531 after epoch 45.
|
||||
45 3.51509 4.11214
|
||||
46 3.51267 4.10466
|
||||
47 3.51021 4.09579
|
||||
48 3.50794 4.08656
|
||||
Step size increases to 0.031384 after epoch 49.
|
||||
49 3.50585 4.07728
|
||||
50 3.50391 4.06815
|
||||
51 3.50193 4.05832
|
||||
52 3.50009 4.04887
|
||||
Step size increases to 0.034523 after epoch 53.
|
||||
53 3.49836 4.03931
|
||||
54 3.4968 4.03148
|
||||
55 3.49646 4.01735
|
||||
56 3.49577 4.01706
|
||||
Step size increases to 0.037975 after epoch 57.
|
||||
57 3.49528 4.00918
|
||||
58 3.4945 4.01098
|
||||
59 3.4945 4.0023
|
||||
60 3.49332 4.00347
|
||||
61 3.49319 3.99792
|
||||
62 3.49216 3.99805
|
||||
Step size increases to 0.041772 after epoch 63.
|
||||
63 3.49194 3.99353
|
||||
64 3.49123 3.99357
|
||||
65 3.49167 3.98955
|
||||
66 3.49062 3.98891
|
||||
67 3.49093 3.98663
|
||||
Step size decreases to 0.037595 after epoch 68.
|
||||
68 3.48995 3.98598
|
||||
69 3.49029 3.98279
|
||||
70 3.4883 3.98284
|
||||
71 3.48954 3.98056
|
||||
Step size decreases to 0.033836 after epoch 72.
|
||||
72 3.48769 3.98048
|
||||
73 3.48891 3.97664
|
||||
74 3.48635 3.97757
|
||||
75 3.4881 3.97462
|
||||
Step size decreases to 0.030452 after epoch 76.
|
||||
76 3.48584 3.97574
|
||||
77 3.48749 3.9711
|
||||
78 3.48473 3.97298
|
||||
79 3.4867 3.96937
|
||||
Step size decreases to 0.027407 after epoch 80.
|
||||
80 3.48431 3.97148
|
||||
81 3.48615 3.96648
|
||||
82 3.48339 3.96895
|
||||
83 3.48544 3.96508
|
||||
Step size decreases to 0.024666 after epoch 84.
|
||||
84 3.48307 3.96771
|
||||
85 3.48498 3.96281
|
||||
86 3.48233 3.96547
|
||||
87 3.48437 3.96168
|
||||
Step size decreases to 0.022200 after epoch 88.
|
||||
88 3.4821 3.96443
|
||||
89 3.484 3.95987
|
||||
90 3.48151 3.96249
|
||||
91 3.4835 3.95894
|
||||
Step size decreases to 0.019980 after epoch 92.
|
||||
92 3.48135 3.96161
|
||||
93 3.48319 3.95745
|
||||
94 3.4809 3.95993
|
||||
95 3.48275 3.95666
|
||||
Step size decreases to 0.017982 after epoch 96.
|
||||
96 3.48078 3.95917
|
||||
97 3.48248 3.95537
|
||||
98 3.48043 3.9577
|
||||
99 3.48207 3.95468
|
||||
Step size decreases to 0.016184 after epoch 100.
|
||||
100 3.48033 3.95705
|
||||
|
||||
Designated epoch number reached. ANFIS training completed at epoch 100.
|
||||
|
||||
Minimal training RMSE = 3.48033
|
||||
Minimal checking RMSE = 3.95468
|
||||
MSE : 16.9264
|
||||
RMSE: 4.11417
|
||||
R2 : 0.664677
|
||||
NMSE: 0.335323
|
||||
NDEI: 0.57907
|
||||
|
||||
|
||||
Model 2: TSK0-3MF (gbellmf × 3 per input)
|
||||
=====================================
|
||||
Initial rules: 243
|
||||
|
||||
ANFIS info:
|
||||
Number of nodes: 524
|
||||
Number of linear parameters: 243
|
||||
Number of nonlinear parameters: 45
|
||||
Total number of parameters: 288
|
||||
Number of training data pairs: 902
|
||||
Number of checking data pairs: 301
|
||||
Number of fuzzy rules: 243
|
||||
|
||||
|
||||
Start training ANFIS ...
|
||||
|
||||
1 3.48949 6.18916
|
||||
2 3.45915 6.06247
|
||||
3 3.41872 5.91346
|
||||
4 3.38236 5.73296
|
||||
Step size increases to 0.011000 after epoch 5.
|
||||
5 3.353 5.57846
|
||||
6 3.32197 5.47698
|
||||
7 3.28593 5.3751
|
||||
8 3.2483 5.26731
|
||||
Step size increases to 0.012100 after epoch 9.
|
||||
9 3.20873 5.1596
|
||||
10 3.16709 5.06056
|
||||
11 3.11912 4.97062
|
||||
12 3.07004 4.90448
|
||||
Step size increases to 0.013310 after epoch 13.
|
||||
13 3.02187 4.86429
|
||||
14 2.97737 4.85458
|
||||
15 2.93582 4.88936
|
||||
16 2.90427 4.98087
|
||||
Step size increases to 0.014641 after epoch 17.
|
||||
17 2.88219 5.12211
|
||||
18 2.86471 5.26311
|
||||
19 2.84624 5.36588
|
||||
20 2.82829 5.46658
|
||||
Step size increases to 0.016105 after epoch 21.
|
||||
21 2.81115 5.55392
|
||||
22 2.79487 5.67287
|
||||
23 2.77853 5.64686
|
||||
24 2.76872 6.11556
|
||||
25 2.77206 5.14558
|
||||
26 2.75702 6.19896
|
||||
27 2.7602 5.14994
|
||||
Step size decreases to 0.014495 after epoch 28.
|
||||
28 2.74638 6.24052
|
||||
29 2.75098 5.11992
|
||||
30 2.73409 6.17768
|
||||
31 2.74283 5.10258
|
||||
Step size decreases to 0.013045 after epoch 32.
|
||||
32 2.72821 6.17703
|
||||
33 2.73726 5.07044
|
||||
34 2.7204 6.07671
|
||||
35 2.73153 5.05552
|
||||
Step size decreases to 0.011741 after epoch 36.
|
||||
36 2.7168 6.06383
|
||||
37 2.7272 5.03036
|
||||
38 2.71115 5.95635
|
||||
39 2.72224 5.02069
|
||||
Step size decreases to 0.010567 after epoch 40.
|
||||
40 2.70844 5.94394
|
||||
41 2.71842 5.00057
|
||||
42 2.70399 5.83984
|
||||
43 2.71392 4.99542
|
||||
Step size decreases to 0.009510 after epoch 44.
|
||||
44 2.70177 5.83091
|
||||
45 2.71053 4.9791
|
||||
46 2.69812 5.73434
|
||||
47 2.70652 4.97696
|
||||
Step size decreases to 0.008559 after epoch 48.
|
||||
48 2.69623 5.72879
|
||||
49 2.70354 4.96362
|
||||
50 2.69316 5.64101
|
||||
51 2.70004 4.96338
|
||||
Step size decreases to 0.007703 after epoch 52.
|
||||
52 2.69153 5.63813
|
||||
53 2.69744 4.95251
|
||||
54 2.68891 5.55924
|
||||
55 2.69442 4.9536
|
||||
Step size decreases to 0.006933 after epoch 56.
|
||||
56 2.68746 5.55841
|
||||
57 2.69215 4.94479
|
||||
58 2.6852 5.48794
|
||||
59 2.68954 4.94684
|
||||
Step size decreases to 0.006239 after epoch 60.
|
||||
60 2.68391 5.48866
|
||||
61 2.68755 4.93971
|
||||
62 2.68193 5.42588
|
||||
63 2.68528 4.94243
|
||||
Step size decreases to 0.005616 after epoch 64.
|
||||
64 2.68076 5.42767
|
||||
65 2.68353 4.93659
|
||||
66 2.67901 5.37175
|
||||
67 2.68155 4.93972
|
||||
Step size decreases to 0.005054 after epoch 68.
|
||||
68 2.67793 5.37416
|
||||
69 2.68 4.93481
|
||||
70 2.67637 5.32424
|
||||
71 2.67826 4.93813
|
||||
Step size decreases to 0.004549 after epoch 72.
|
||||
72 2.67537 5.32691
|
||||
73 2.67687 4.93383
|
||||
74 2.67396 5.28218
|
||||
75 2.67533 4.93716
|
||||
Step size decreases to 0.004094 after epoch 76.
|
||||
76 2.67302 5.28479
|
||||
77 2.67407 4.93317
|
||||
78 2.67174 5.24452
|
||||
79 2.6727 4.93637
|
||||
Step size decreases to 0.003684 after epoch 80.
|
||||
80 2.67084 5.24682
|
||||
81 2.67156 4.93247
|
||||
82 2.66967 5.21037
|
||||
83 2.67031 4.93543
|
||||
Step size decreases to 0.003316 after epoch 84.
|
||||
84 2.66881 5.21218
|
||||
85 2.66927 4.93142
|
||||
86 2.66772 5.17901
|
||||
87 2.66813 4.93412
|
||||
Step size decreases to 0.002984 after epoch 88.
|
||||
88 2.66689 5.1802
|
||||
89 2.66716 4.92983
|
||||
90 2.66588 5.14986
|
||||
91 2.66611 4.93226
|
||||
Step size decreases to 0.002686 after epoch 92.
|
||||
92 2.66507 5.15035
|
||||
93 2.6652 4.92758
|
||||
94 2.66412 5.12247
|
||||
95 2.66422 4.92977
|
||||
Step size decreases to 0.002417 after epoch 96.
|
||||
96 2.66333 5.1222
|
||||
97 2.66335 4.92461
|
||||
98 2.66244 5.09648
|
||||
99 2.66243 4.92661
|
||||
100 2.66165 5.09539
|
||||
|
||||
Designated epoch number reached. ANFIS training completed at epoch 100.
|
||||
|
||||
Minimal training RMSE = 2.66165
|
||||
Minimal checking RMSE = 4.85458
|
||||
MSE : 17.1406
|
||||
RMSE: 4.14012
|
||||
R2 : 0.660434
|
||||
NMSE: 0.339566
|
||||
NDEI: 0.582723
|
||||
|
||||
|
||||
Model 3: TSK1-2MF (gbellmf × 2 per input)
|
||||
=====================================
|
||||
Initial rules: 32
|
||||
|
||||
ANFIS info:
|
||||
Number of nodes: 92
|
||||
Number of linear parameters: 192
|
||||
Number of nonlinear parameters: 30
|
||||
Total number of parameters: 222
|
||||
Number of training data pairs: 902
|
||||
Number of checking data pairs: 301
|
||||
Number of fuzzy rules: 32
|
||||
|
||||
|
||||
Start training ANFIS ...
|
||||
|
||||
1 2.85009 4.69235
|
||||
2 2.83691 4.61659
|
||||
3 2.82381 4.5385
|
||||
4 2.81077 4.45873
|
||||
Step size increases to 0.011000 after epoch 5.
|
||||
5 2.79775 4.37785
|
||||
6 2.78474 4.29631
|
||||
7 2.7704 4.2064
|
||||
8 2.756 4.11658
|
||||
Step size increases to 0.012100 after epoch 9.
|
||||
9 2.7415 4.02712
|
||||
10 2.72683 3.93818
|
||||
11 2.71042 3.84107
|
||||
12 2.69361 3.74462
|
||||
Step size increases to 0.013310 after epoch 13.
|
||||
13 2.67624 3.64883
|
||||
14 2.65807 3.5539
|
||||
15 2.63684 3.45151
|
||||
16 2.6138 3.35379
|
||||
Step size increases to 0.014641 after epoch 17.
|
||||
17 2.5883 3.2663
|
||||
18 2.55954 3.19899
|
||||
19 2.52297 3.16663
|
||||
20 2.47983 3.20424
|
||||
Step size increases to 0.016105 after epoch 21.
|
||||
21 2.42917 3.3412
|
||||
22 2.37416 3.59905
|
||||
23 2.34422 3.89625
|
||||
24 2.36179 3.60671
|
||||
25 2.33613 3.94345
|
||||
26 2.3545 3.59925
|
||||
Step size decreases to 0.014495 after epoch 27.
|
||||
27 2.32932 3.98109
|
||||
28 2.34864 3.58976
|
||||
29 2.31899 4.00715
|
||||
30 2.34073 3.59264
|
||||
Step size decreases to 0.013045 after epoch 31.
|
||||
31 2.3134 4.05635
|
||||
32 2.33536 3.58825
|
||||
33 2.30527 4.05997
|
||||
34 2.32785 3.59718
|
||||
Step size decreases to 0.011741 after epoch 35.
|
||||
35 2.30083 4.10646
|
||||
36 2.32309 3.59599
|
||||
37 2.29443 4.0886
|
||||
38 2.31615 3.60889
|
||||
Step size decreases to 0.010567 after epoch 39.
|
||||
39 2.29086 4.12772
|
||||
40 2.31186 3.61014
|
||||
41 2.2857 4.09768
|
||||
42 2.30547 3.62593
|
||||
Step size decreases to 0.009510 after epoch 43.
|
||||
43 2.28273 4.13163
|
||||
44 2.30157 3.62894
|
||||
45 2.27844 4.09721
|
||||
46 2.29579 3.64627
|
||||
Step size decreases to 0.008559 after epoch 47.
|
||||
47 2.27589 4.1275
|
||||
48 2.29224 3.65042
|
||||
49 2.27223 4.09275
|
||||
50 2.2871 3.6682
|
||||
Step size decreases to 0.007703 after epoch 51.
|
||||
51 2.27 4.11992
|
||||
52 2.28389 3.67307
|
||||
53 2.26682 4.08658
|
||||
54 2.27936 3.69061
|
||||
Step size decreases to 0.006933 after epoch 55.
|
||||
55 2.26483 4.11089
|
||||
56 2.27645 3.6959
|
||||
57 2.26204 4.07957
|
||||
58 2.27247 3.7128
|
||||
Step size decreases to 0.006239 after epoch 59.
|
||||
59 2.26024 4.10119
|
||||
60 2.26983 3.71823
|
||||
61 2.25776 4.07206
|
||||
62 2.26633 3.73423
|
||||
Step size decreases to 0.005616 after epoch 63.
|
||||
63 2.25611 4.09117
|
||||
64 2.26395 3.73959
|
||||
65 2.25389 4.06416
|
||||
66 2.26087 3.75453
|
||||
Step size decreases to 0.005054 after epoch 67.
|
||||
67 2.25236 4.08094
|
||||
68 2.25871 3.75961
|
||||
69 2.25036 4.0559
|
||||
70 2.25599 3.77339
|
||||
Step size decreases to 0.004549 after epoch 71.
|
||||
71 2.24893 4.07052
|
||||
72 2.25403 3.77802
|
||||
73 2.24711 4.04729
|
||||
74 2.25162 3.7906
|
||||
Step size decreases to 0.004094 after epoch 75.
|
||||
75 2.24576 4.05991
|
||||
76 2.24983 3.79463
|
||||
77 2.24409 4.03828
|
||||
78 2.24767 3.806
|
||||
Step size decreases to 0.003684 after epoch 79.
|
||||
79 2.2428 4.04906
|
||||
80 2.24604 3.80931
|
||||
81 2.24127 4.02885
|
||||
82 2.2441 3.81953
|
||||
Step size decreases to 0.003316 after epoch 83.
|
||||
83 2.24002 4.03794
|
||||
84 2.2426 3.82204
|
||||
85 2.2386 4.019
|
||||
86 2.24084 3.83118
|
||||
Step size decreases to 0.002984 after epoch 87.
|
||||
87 2.2374 4.02654
|
||||
88 2.23945 3.83283
|
||||
89 2.23608 4.00874
|
||||
90 2.23783 3.84101
|
||||
Step size decreases to 0.002686 after epoch 91.
|
||||
91 2.23491 4.01486
|
||||
92 2.23654 3.84179
|
||||
93 2.23367 3.99812
|
||||
94 2.23504 3.84916
|
||||
Step size decreases to 0.002417 after epoch 95.
|
||||
95 2.23253 4.00297
|
||||
96 2.23383 3.84907
|
||||
97 2.23136 3.98724
|
||||
98 2.23243 3.8558
|
||||
Step size decreases to 0.002176 after epoch 99.
|
||||
99 2.23025 3.99091
|
||||
100 2.23128 3.85487
|
||||
|
||||
Designated epoch number reached. ANFIS training completed at epoch 100.
|
||||
|
||||
Minimal training RMSE = 2.23025
|
||||
Minimal checking RMSE = 3.16663
|
||||
MSE : 12.8953
|
||||
RMSE: 3.591
|
||||
R2 : 0.744537
|
||||
NMSE: 0.255463
|
||||
NDEI: 0.505433
|
||||
|
||||
|
||||
Model 4: TSK1-3MF (gbellmf × 3 per input)
|
||||
=====================================
|
||||
Initial rules: 243
|
||||
|
||||
ANFIS info:
|
||||
Number of nodes: 524
|
||||
Number of linear parameters: 1458
|
||||
Number of nonlinear parameters: 45
|
||||
Total number of parameters: 1503
|
||||
Number of training data pairs: 902
|
||||
Number of checking data pairs: 301
|
||||
Number of fuzzy rules: 243
|
||||
|
||||
Warning: Number of training data is smaller than number of modifiable parameters.
|
||||
> In anfis>trainFIS (line 203)
|
||||
In anfis>anfisWithOptionObject (line 109)
|
||||
In anfis (line 68)
|
||||
In scenario1 (line 66)
|
||||
|
||||
Start training ANFIS ...
|
||||
|
||||
1 2.18266 13.9384
|
||||
2 2.16266 14.302
|
||||
3 2.14218 14.771
|
||||
4 2.12092 15.3612
|
||||
Step size increases to 0.011000 after epoch 5.
|
||||
5 2.0985 16.0849
|
||||
6 2.07446 16.9408
|
||||
7 2.04563 17.9967
|
||||
8 2.01387 19.1103
|
||||
Step size increases to 0.012100 after epoch 9.
|
||||
9 1.97901 20.1856
|
||||
10 1.94113 21.1309
|
||||
11 1.89631 21.9439
|
||||
12 1.84864 22.4876
|
||||
Step size increases to 0.013310 after epoch 13.
|
||||
13 1.79858 22.7812
|
||||
14 1.74656 22.8849
|
||||
15 1.68767 22.8664
|
||||
16 1.62771 22.7684
|
||||
Step size increases to 0.014641 after epoch 17.
|
||||
17 1.56748 22.5829
|
||||
18 1.50801 22.2279
|
||||
19 1.44544 21.5009
|
||||
20 1.38899 20.2966
|
||||
Step size increases to 0.016105 after epoch 21.
|
||||
21 1.34211 18.5912
|
||||
22 1.30602 16.488
|
||||
23 1.27558 14.1361
|
||||
24 1.25531 13.5366
|
||||
Step size increases to 0.017716 after epoch 25.
|
||||
25 1.22833 11.9472
|
||||
26 1.20732 11.145
|
||||
27 1.17735 9.2776
|
||||
28 1.15854 8.56663
|
||||
Step size increases to 0.019487 after epoch 29.
|
||||
29 1.12679 6.41978
|
||||
30 1.12503 6.60586
|
||||
31 1.09297 4.9284
|
||||
32 1.10129 5.326
|
||||
33 1.06957 4.68211
|
||||
34 1.08124 4.74092
|
||||
Step size decreases to 0.017538 after epoch 35.
|
||||
35 1.05297 4.99736
|
||||
36 1.0659 4.73721
|
||||
37 1.03553 5.54205
|
||||
38 1.04981 5.0522
|
||||
Step size decreases to 0.015785 after epoch 39.
|
||||
39 1.02529 6.28831
|
||||
40 1.03688 5.44902
|
||||
41 1.009 7.02564
|
||||
42 1.02048 5.91912
|
||||
Step size decreases to 0.014206 after epoch 43.
|
||||
43 0.998267 8.03611
|
||||
44 1.00675 6.31568
|
||||
45 0.981053 8.87018
|
||||
46 0.992248 6.70868
|
||||
Step size decreases to 0.012786 after epoch 47.
|
||||
47 0.973147 9.9055
|
||||
48 0.983062 6.97916
|
||||
49 0.958957 10.505
|
||||
50 0.973683 7.26996
|
||||
Step size decreases to 0.011507 after epoch 51.
|
||||
51 0.954367 11.3465
|
||||
52 0.967501 7.4856
|
||||
53 0.942356 11.7345
|
||||
54 0.960308 7.75761
|
||||
Step size decreases to 0.010356 after epoch 55.
|
||||
55 0.939278 12.4896
|
||||
56 0.955315 7.97285
|
||||
57 0.928728 12.7622
|
||||
58 0.948976 8.26813
|
||||
Step size decreases to 0.009321 after epoch 59.
|
||||
59 0.926441 13.4981
|
||||
60 0.944515 8.50717
|
||||
61 0.91711 13.7127
|
||||
62 0.938576 8.84703
|
||||
Step size decreases to 0.008389 after epoch 63.
|
||||
63 0.915346 14.4517
|
||||
64 0.934437 9.12183
|
||||
65 0.907143 14.6391
|
||||
66 0.928786 9.51736
|
||||
Step size decreases to 0.007550 after epoch 67.
|
||||
67 0.905787 15.3775
|
||||
68 0.924938 9.83222
|
||||
69 0.898656 15.5517
|
||||
70 0.919613 10.2861
|
||||
Step size decreases to 0.006795 after epoch 71.
|
||||
71 0.89764 16.2705
|
||||
72 0.916096 10.6381
|
||||
73 0.891518 16.4343
|
||||
74 0.911178 11.1444
|
||||
Step size decreases to 0.006115 after epoch 75.
|
||||
75 0.890784 17.1082
|
||||
76 0.908033 11.5239
|
||||
77 0.885587 17.2586
|
||||
78 0.903579 12.0686
|
||||
Step size decreases to 0.005504 after epoch 79.
|
||||
79 0.885076 17.8654
|
||||
80 0.90081 12.4607
|
||||
81 0.8807 17.9986
|
||||
82 0.896814 13.0247
|
||||
Step size decreases to 0.004953 after epoch 83.
|
||||
83 0.880351 18.5253
|
||||
84 0.894379 13.4131
|
||||
85 0.876676 18.6391
|
||||
86 0.890789 13.9761
|
||||
Step size decreases to 0.004458 after epoch 87.
|
||||
87 0.876438 19.0838
|
||||
88 0.888624 14.3467
|
||||
89 0.873343 19.1782
|
||||
90 0.885379 14.8903
|
||||
Step size decreases to 0.004012 after epoch 91.
|
||||
91 0.873172 19.5467
|
||||
92 0.883427 15.2336
|
||||
93 0.870543 19.6236
|
||||
94 0.880489 15.7433
|
||||
Step size decreases to 0.003611 after epoch 95.
|
||||
95 0.870406 19.9261
|
||||
96 0.878716 16.0547
|
||||
97 0.868147 19.9882
|
||||
98 0.876081 16.5202
|
||||
Step size decreases to 0.003250 after epoch 99.
|
||||
99 0.868021 20.2362
|
||||
100 0.874473 16.7991
|
||||
|
||||
Designated epoch number reached. ANFIS training completed at epoch 100.
|
||||
|
||||
Minimal training RMSE = 0.868021
|
||||
Minimal checking RMSE = 4.68211
|
||||
MSE : 30.4753
|
||||
RMSE: 5.52045
|
||||
R2 : 0.396266
|
||||
NMSE: 0.603734
|
||||
NDEI: 0.777003
|
||||
|
||||
Metrics Model_1 Model_2 Model_3 Model_4
|
||||
_______ _______ _______ _______ _______
|
||||
|
||||
MSE 16.926 17.141 12.895 30.475
|
||||
RMSE 4.1142 4.1401 3.591 5.5204
|
||||
R2 0.66468 0.66043 0.74454 0.39627
|
||||
NMSE 0.33532 0.33957 0.25546 0.60373
|
||||
NDEI 0.57907 0.58272 0.50543 0.777
|
||||
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,17 @@
|
||||
function [mse, rmse, r2, nmse, ndei] = evaluate(pred, truth)
|
||||
%EVALUATE Summary of this function goes here
|
||||
% Detailed explanation goes here
|
||||
|
||||
% MSE / RMSE
|
||||
mse = mean((truth - pred).^2);
|
||||
rmse = sqrt(mse);
|
||||
|
||||
% R^2 - NMSE - NDEI
|
||||
ss_res = sum((truth - pred).^2);
|
||||
ss_tot = sum((truth - mean(truth)).^2);
|
||||
|
||||
nmse = ss_res / ss_tot;
|
||||
ndei = sqrt(nmse);
|
||||
r2 = 1 - nmse;
|
||||
|
||||
end
|
||||
Binary file not shown.
|
After Width: | Height: | Size: 76 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 48 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 133 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 57 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 76 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 68 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 154 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 55 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 74 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 66 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 125 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 54 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 52 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 68 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 154 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 48 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 49 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 45 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 47 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 67 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 63 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 308 KiB |
Binary file not shown.
|
After Width: | Height: | Size: 64 KiB |
@@ -0,0 +1,148 @@
|
||||
function plot_results1(init_fis, final_fis, trn_error, val_error, y_true, y_pred, model_id)
|
||||
%PLOT_RESULTS
|
||||
% Creates and saves:
|
||||
% (A) One consolidated figure for membership functions (all inputs):
|
||||
% - 2 x Ninputs tiled layout
|
||||
% - Top row: BEFORE training (all MFs per input in one subplot)
|
||||
% - Bottom row: AFTER training (all MFs per input in one subplot)
|
||||
% (B) Learning curves (train & validation)
|
||||
% (C) Predicted vs Actual (with y=x reference)
|
||||
% (D) Prediction Error (time series + MAE)
|
||||
%
|
||||
% All figures are saved to ./figures as both PDF and PNG.
|
||||
%
|
||||
% Usage:
|
||||
% plot_results(init_fis, final_fis, trn_error, val_error, y_true, y_pred, model_id)
|
||||
|
||||
% Input checks
|
||||
if nargin < 7
|
||||
error('plot_results: not enough inputs. Expected 7 arguments.');
|
||||
end
|
||||
if numel(y_true) ~= numel(y_pred)
|
||||
error('plot_results: y_true and y_pred must have the same length.');
|
||||
end
|
||||
y_true = y_true(:);
|
||||
y_pred = y_pred(:);
|
||||
|
||||
% Output directory
|
||||
outdir = fullfile('.', 'figures_scn1');
|
||||
if ~exist(outdir, 'dir'); mkdir(outdir); end
|
||||
|
||||
% (A) MEMBERSHIP FUNCTIONS — ONE FIGURE, 2 x Ninputs (BEFORE / AFTER)
|
||||
% =====================================================================
|
||||
nInputs = numel(init_fis.Inputs);
|
||||
|
||||
% Precompute x/y for all inputs to keep consistent axes
|
||||
Xb = cell(1, nInputs); Yb = cell(1, nInputs); % BEFORE
|
||||
Xa = cell(1, nInputs); Ya = cell(1, nInputs); % AFTER
|
||||
nMFb = zeros(1, nInputs); nMFa = zeros(1, nInputs);
|
||||
|
||||
for inIdx = 1:nInputs
|
||||
[xb, yb] = plotmf(init_fis, 'input', inIdx);
|
||||
[xa, ya] = plotmf(final_fis, 'input', inIdx);
|
||||
Xb{inIdx} = xb; Yb{inIdx} = yb; nMFb(inIdx) = size(yb,2);
|
||||
Xa{inIdx} = xa; Ya{inIdx} = ya; nMFa(inIdx) = size(ya,2);
|
||||
end
|
||||
|
||||
% Create consolidated figure
|
||||
f = figure('Name', sprintf('MFs — Model %d', model_id), 'Color','w');
|
||||
tl = tiledlayout(f, 2, nInputs, 'TileSpacing','compact', 'Padding','compact');
|
||||
|
||||
% Titles for rows
|
||||
rowTitles = {'BEFORE training', 'AFTER training'};
|
||||
|
||||
for inIdx = 1:nInputs
|
||||
% BEFORE (top row)
|
||||
nexttile(tl, inIdx); % row 1, col inIdx
|
||||
hold on; grid on;
|
||||
xb = Xb{inIdx}; yb = Yb{inIdx}; kB = size(yb,2);
|
||||
for k = 1:kB
|
||||
plot(xb, yb(:,k), 'LineWidth', 1.5);
|
||||
end
|
||||
xlabel(sprintf('Input x%d', inIdx), 'Interpreter','latex');
|
||||
ylabel('Membership', 'Interpreter','latex');
|
||||
title(sprintf('%s — x%d', rowTitles{1}, inIdx), 'Interpreter','latex');
|
||||
|
||||
% Optional legend only on the first tile to reduce clutter
|
||||
if inIdx == 1
|
||||
lgdB = arrayfun(@(k) sprintf('MF %d', k), 1:kB, 'UniformOutput', false);
|
||||
legend(lgdB, 'Location','best');
|
||||
end
|
||||
|
||||
% AFTER (bottom row)
|
||||
nexttile(tl, nInputs + inIdx); % row 2, col inIdx
|
||||
hold on; grid on;
|
||||
xa = Xa{inIdx}; ya = Ya{inIdx}; kA = size(ya,2);
|
||||
for k = 1:kA
|
||||
plot(xa, ya(:,k), 'LineWidth', 1.5);
|
||||
end
|
||||
xlabel(sprintf('Input x%d', inIdx), 'Interpreter','latex');
|
||||
ylabel('Membership', 'Interpreter','latex');
|
||||
title(sprintf('%s — x%d', rowTitles{2}, inIdx), 'Interpreter','latex');
|
||||
|
||||
if inIdx == 1
|
||||
lgdA = arrayfun(@(k) sprintf('MF %d', k), 1:kA, 'UniformOutput', false);
|
||||
legend(lgdA, 'Location','best');
|
||||
end
|
||||
end
|
||||
|
||||
% Super-title
|
||||
sgtitle(tl, sprintf('Membership Functions — Model %d', model_id), 'Interpreter','latex');
|
||||
|
||||
% Save consolidated MF figure
|
||||
save_figure(f, fullfile(outdir, sprintf('model_%d_mfs_all_inputs', model_id)));
|
||||
|
||||
% (B) LEARNING CURVES (train & validation)
|
||||
% ============================================
|
||||
f = figure('Name', sprintf('Learning Curves - Model %d', model_id), 'Color','w');
|
||||
plot(trn_error, 'LineWidth', 1.5); hold on; grid on;
|
||||
if ~isempty(val_error)
|
||||
plot(val_error, 'LineWidth', 1.5);
|
||||
legend('Train','Validation','Location','best');
|
||||
else
|
||||
legend('Train','Location','best');
|
||||
end
|
||||
xlabel('Epoch', 'Interpreter','latex');
|
||||
ylabel('Error', 'Interpreter','latex');
|
||||
title(sprintf('Learning Curves | Model %d', model_id), 'Interpreter','latex');
|
||||
save_figure(f, fullfile(outdir, sprintf('model_%d_learning_curves', model_id)));
|
||||
|
||||
% (C) PREDICTED vs ACTUAL
|
||||
% ============================================
|
||||
f = figure('Name', sprintf('Predicted vs Actual - Model %d', model_id), 'Color','w');
|
||||
plot(y_true, y_pred, '.', 'MarkerSize', 10); hold on; grid on;
|
||||
|
||||
mins = min([y_true; y_pred]);
|
||||
maxs = max([y_true; y_pred]);
|
||||
plot([mins maxs], [mins maxs], 'k-', 'LineWidth', 1);
|
||||
|
||||
xlabel('Actual', 'Interpreter','latex');
|
||||
ylabel('Predicted', 'Interpreter','latex');
|
||||
title(sprintf('Predicted vs Actual | Model %d', model_id), 'Interpreter','latex');
|
||||
save_figure(f, fullfile(outdir, sprintf('model_%d_pred_vs_actual', model_id)));
|
||||
|
||||
% (D) PREDICTION ERROR
|
||||
% ============================================
|
||||
err = y_true - y_pred;
|
||||
f = figure('Name', sprintf('Prediction Error - Model %d', model_id), 'Color','w');
|
||||
plot(err, 'k'); grid on;
|
||||
xlabel('Iteration', 'Interpreter','latex');
|
||||
ylabel('Prediction Error', 'Interpreter','latex');
|
||||
title('Prediction Error', 'Interpreter','latex');
|
||||
mae = mean(abs(err));
|
||||
try
|
||||
subtitle(sprintf('Mean absolute error: %f', mae), 'Interpreter','latex');
|
||||
catch
|
||||
title(sprintf('Prediction Error (MAE=%.6f)', mae), 'Interpreter','latex');
|
||||
end
|
||||
save_figure(f, fullfile(outdir, sprintf('model_%d_error', model_id)));
|
||||
end
|
||||
|
||||
% Helper: consistent saving (PDF + PNG, landscape)
|
||||
function save_figure(fig_handle, basepath)
|
||||
set(fig_handle, 'PaperUnits','normalized');
|
||||
set(fig_handle, 'PaperPosition', [0 0 1 1]);
|
||||
set(fig_handle, 'PaperOrientation', 'landscape');
|
||||
% print(fig_handle, '-dpdf', [basepath '.pdf']);
|
||||
print(fig_handle, '-dpng', [basepath '.png']);
|
||||
end
|
||||
@@ -0,0 +1,107 @@
|
||||
function plot_results2(init_fis, final_fis, trn_error, val_error, y_true, y_pred, ...
|
||||
feature_grid, radius_grid, cv_scores, cv_rules, sel_idx)
|
||||
% PLOTS FOR SCENARIO 2:
|
||||
% (1) CV heatmap (mean CV error over grid) with overlaid #rules
|
||||
% (2) Error vs rules (aggregated) & Error vs #features (best over radii)
|
||||
% (3) Final model: Learning curves, Predicted vs Actual, Residuals
|
||||
% (4) MFs (subset of selected inputs) BEFORE/AFTER
|
||||
%
|
||||
% All figures are saved under ./figures
|
||||
|
||||
outdir = fullfile('.', 'figures_scn2'); if ~exist(outdir,'dir'), mkdir(outdir); end
|
||||
|
||||
% (1) CV heatmap
|
||||
f = figure('Color','w','Name','CV Heatmap');
|
||||
imagesc(radius_grid, feature_grid, cv_scores);
|
||||
set(gca,'YDir','normal'); xlabel('Cluster radius r_\alpha','Interpreter','latex');
|
||||
ylabel('#Features','Interpreter','latex');
|
||||
title('Mean CV Error','Interpreter','latex'); colorbar; grid on;
|
||||
% overlay mean rules
|
||||
hold on;
|
||||
for i=1:numel(feature_grid)
|
||||
for j=1:numel(radius_grid)
|
||||
text(radius_grid(j), feature_grid(i), sprintf('%d', cv_rules(i,j)), ...
|
||||
'HorizontalAlignment','center','Color','w','FontSize',8);
|
||||
end
|
||||
end
|
||||
try
|
||||
subtitle('Numbers show mean #rules','Interpreter','latex');
|
||||
end
|
||||
print(f, fullfile(outdir,'scn2_cv_heatmap'), '-dpng');
|
||||
|
||||
% (2a) Error vs #rules (aggregate by identical rule-counts)
|
||||
uniq_rules = unique(cv_rules(:));
|
||||
err_vs_rules = zeros(size(uniq_rules));
|
||||
for k=1:numel(uniq_rules)
|
||||
err_vs_rules(k) = mean(cv_scores(cv_rules == uniq_rules(k)));
|
||||
end
|
||||
f = figure('Color','w','Name','Error vs Rules');
|
||||
plot(uniq_rules, err_vs_rules, 'o-','LineWidth',1.5); grid on;
|
||||
xlabel('#Rules','Interpreter','latex'); ylabel('Mean CV Error','Interpreter','latex');
|
||||
title('Error vs Rules','Interpreter','latex');
|
||||
print(f, fullfile(outdir,'scn2_error_vs_rules'), '-dpng');
|
||||
|
||||
% (2b) Best CV error vs #features (min across radii)
|
||||
[min_err_per_feat,~] = min(cv_scores,[],2);
|
||||
f = figure('Color','w','Name','Error vs Features');
|
||||
plot(feature_grid, min_err_per_feat, 's-','LineWidth',1.5); grid on;
|
||||
xlabel('#Features','Interpreter','latex'); ylabel('Best CV Error','Interpreter','latex');
|
||||
title('Best CV Error vs #Features','Interpreter','latex');
|
||||
print(f, fullfile(outdir,'scn2_error_vs_features'), '-dpng');
|
||||
|
||||
% (3a) Learning curves
|
||||
f = figure('Color','w','Name','Learning Curves');
|
||||
plot(trn_error,'LineWidth',1.5); hold on; grid on;
|
||||
if ~isempty(val_error)
|
||||
plot(val_error,'LineWidth',1.5); legend('Train','Validation','Location','best');
|
||||
else
|
||||
legend('Train','Location','best');
|
||||
end
|
||||
xlabel('Epoch','Interpreter','latex'); ylabel('Error','Interpreter','latex');
|
||||
title('Learning Curves','Interpreter','latex');
|
||||
print(f, fullfile(outdir,'scn2_final_learning_curves'), '-dpng');
|
||||
|
||||
% (3b) Predicted vs Actual
|
||||
y_true = y_true(:); y_pred = y_pred(:);
|
||||
f = figure('Color','w','Name','Predicted vs Actual');
|
||||
plot(y_true, y_pred, '.', 'MarkerSize', 10); hold on; grid on;
|
||||
mins = min([y_true; y_pred]); maxs = max([y_true; y_pred]);
|
||||
plot([mins maxs], [mins maxs], 'k-', 'LineWidth', 1);
|
||||
xlabel('Actual','Interpreter','latex'); ylabel('Predicted','Interpreter','latex');
|
||||
title('Predicted vs Actual','Interpreter','latex');
|
||||
print(f, fullfile(outdir,'scn2_final_pred_vs_actual'), '-dpng');
|
||||
|
||||
% (3c) Residuals (time series)
|
||||
err = y_true - y_pred;
|
||||
f = figure('Color','w','Name','Prediction Error');
|
||||
plot(err, 'k'); grid on;
|
||||
xlabel('Sample','Interpreter','latex'); ylabel('Error','Interpreter','latex');
|
||||
title('Prediction Error','Interpreter','latex');
|
||||
mae = mean(abs(err));
|
||||
try
|
||||
subtitle(sprintf('Mean absolute error: %.6f', mae), 'Interpreter','latex');
|
||||
end
|
||||
print(f, fullfile(outdir,'scn2_final_error_series'), '-dpng');
|
||||
|
||||
% (4) MFs (subset of selected inputs) BEFORE/AFTER
|
||||
% Show up to 3 selected inputs for clarity
|
||||
nShow = min( min(3, numel(sel_idx)), numel(init_fis.Inputs) );
|
||||
if nShow > 0
|
||||
f = figure('Color','w','Name','MFs (subset)');
|
||||
tl = tiledlayout(2, nShow, 'TileSpacing','compact', 'Padding','compact');
|
||||
for k=1:nShow
|
||||
inIdx = k; % first few selected
|
||||
[xb,yb] = plotmf(init_fis,'input',inIdx);
|
||||
[xa,ya] = plotmf(final_fis,'input',inIdx);
|
||||
nexttile(tl,k); hold on; grid on; plot(xb,yb,'LineWidth',1.2);
|
||||
title(sprintf('BEFORE x_{%d}', sel_idx(k)),'Interpreter','latex');
|
||||
xlabel(sprintf('x_{%d}', sel_idx(k)),'Interpreter','latex'); ylabel('Membership','Interpreter','latex');
|
||||
|
||||
nexttile(tl,nShow+k); hold on; grid on; plot(xa,ya,'LineWidth',1.2);
|
||||
title(sprintf('AFTER x_{%d}', sel_idx(k)),'Interpreter','latex');
|
||||
xlabel(sprintf('x_{%d}', sel_idx(k)),'Interpreter','latex'); ylabel('Membership','Interpreter','latex');
|
||||
end
|
||||
sgtitle(tl, 'Membership Functions (subset)','Interpreter','latex');
|
||||
print(f, fullfile(outdir,'scn2_final_mfs_subset'), '-dpng');
|
||||
end
|
||||
end
|
||||
@@ -0,0 +1,39 @@
|
||||
function [X_trn_s, X_val_s, X_chk_s, stats] = preprocess_data(X_trn, X_val, X_chk, mode)
|
||||
%SPLITSET Splits the data-set to train, eval,check data
|
||||
% Randomly split the data set according to ratios
|
||||
%
|
||||
|
||||
if nargin < 4, mode = 1; end % 1: min-max, 2: z-score
|
||||
|
||||
switch mode
|
||||
case 1 % Min–Max to [0,1] using TRAIN stats
|
||||
xmin = min(X_trn,[],1);
|
||||
xmax = max(X_trn,[],1);
|
||||
range = xmax - xmin;
|
||||
range(range==0) = 1;
|
||||
X_trn_s = (X_trn - xmin) ./ range;
|
||||
X_val_s = (X_val - xmin) ./ range;
|
||||
X_chk_s = (X_chk - xmin) ./ range;
|
||||
stats = struct( ...
|
||||
'type', 'minmax', ...
|
||||
'xmin', xmin, ...
|
||||
'xmax', xmax ...
|
||||
);
|
||||
|
||||
case 2 % Z-score using TRAIN stats (chatGPT gave me this one)
|
||||
mu = mean(X_trn,1);
|
||||
sig = std(X_trn,0,1);
|
||||
sig(sig==0) = 1;
|
||||
X_trn_s = (X_trn - mu) ./ sig;
|
||||
X_val_s = (X_val - mu) ./ sig;
|
||||
X_chk_s = (X_chk - mu) ./ sig;
|
||||
stats = struct(...
|
||||
'type', 'zscore', ...
|
||||
'mu', mu, ...
|
||||
'sig', sig ...
|
||||
);
|
||||
|
||||
otherwise
|
||||
error('Unknown mode.');
|
||||
end
|
||||
end
|
||||
@@ -0,0 +1,94 @@
|
||||
%% Scenario1 (TSK - Airfoil Self-Noise)
|
||||
%
|
||||
% Assignment 3 in Fuzzy systems
|
||||
%
|
||||
% author:
|
||||
% Christos Choutouridis ΑΕΜ 8997
|
||||
% cchoutou@ece.auth.gr
|
||||
%
|
||||
clear; clc; close all;
|
||||
|
||||
|
||||
% Configuration
|
||||
% --------------------------------
|
||||
config.mf_types = ["constant", "constant", "linear", "linear"];
|
||||
config.num_mf = [2, 3, 2, 3]; % #MFs per input
|
||||
config.model_names = ["TSK0-2MF", "TSK0-3MF", "TSK1-2MF", "TSK1-3MF"];
|
||||
config.Nepochs = 100;
|
||||
|
||||
rng(42,'twister');
|
||||
|
||||
% Load dataset (Airfoil Self-Noise: 5 inputs, 1 target)
|
||||
data = load("Datasets/airfoil_self_noise.dat");
|
||||
% Split data: 60% train, 20% val, 20% check(test)
|
||||
[X_trn, y_trn, X_val, y_val, X_chk, y_chk] = split_data(data, [0.6 0.2 0.2], 42);
|
||||
% Preprocess inputs (mode=1 -> MinMax to [0,1])
|
||||
[X_trn, X_val, X_chk, stats] = preprocess_data(X_trn, X_val, X_chk, 1);
|
||||
|
||||
% Pack data for anfis/evalfis convenience
|
||||
trn_data = [X_trn, y_trn];
|
||||
val_data = [X_val, y_val];
|
||||
chk_X = X_chk;
|
||||
chk_y = y_chk;
|
||||
|
||||
|
||||
% Results
|
||||
T = table(['MSE '; 'RMSE'; 'R2 '; 'NMSE'; 'NDEI'], 'VariableNames', {'Metrics'});
|
||||
|
||||
all_models = cell(1,4);
|
||||
all_trnErr = cell(1,4);
|
||||
all_valErr = cell(1,4);
|
||||
|
||||
for i = 1:length(config.mf_types)
|
||||
fprintf('\n');
|
||||
fprintf('Model %d: %s (gbellmf × %d per input)\n', i, config.model_names(i), config.num_mf(i));
|
||||
fprintf('=====================================\n');
|
||||
|
||||
% Step 1 - Initial FIS via Grid Partition
|
||||
opt_g = genfisOptions("GridPartition", ...
|
||||
"InputMembershipFunctionType", "gbellmf", ...
|
||||
"NumMembershipFunctions", config.num_mf(i), ...
|
||||
"OutputMembershipFunctionType", config.mf_types(i));
|
||||
|
||||
init_fis = genfis(X_trn, y_trn, opt_g);
|
||||
|
||||
% Info rules and MF
|
||||
nRules = numel(init_fis.Rules);
|
||||
fprintf('Initial rules: %d\n', nRules);
|
||||
|
||||
% Step 2 - Train with ANFIS (Hybrid: BP + LSE)
|
||||
opt_a = anfisOptions( ...
|
||||
"InitialFis", init_fis, ...
|
||||
"ValidationData", val_data, ...
|
||||
"EpochNumber", config.Nepochs, ...
|
||||
"OptimizationMethod", 1 ... % 1=Hybrid
|
||||
);
|
||||
[trn_fis, trn_error, ~, val_fis, val_error] = anfis(trn_data, opt_a);
|
||||
|
||||
final_fis = val_fis; % Select final
|
||||
|
||||
% Step 3 - Evaluate on Check/Test set
|
||||
y_hat = evalfis(final_fis, chk_X);
|
||||
[mse, rmse, r2, nmse, ndei] = evaluate(y_hat, chk_y);
|
||||
|
||||
% Print & store metrics
|
||||
fprintf('MSE : %g\n', mse);
|
||||
fprintf('RMSE: %g\n', rmse);
|
||||
fprintf('R2 : %g\n', r2);
|
||||
fprintf('NMSE: %g\n', nmse);
|
||||
fprintf('NDEI: %g\n', ndei);
|
||||
|
||||
T = addvars(T, [mse; rmse; r2; nmse; ndei], 'NewVariableNames', sprintf('Model_%d',i));
|
||||
|
||||
% Plots & figure exports
|
||||
% - MFs before/after
|
||||
% - Learning curves
|
||||
% - Predicted vs Actual
|
||||
% - Prediction Error
|
||||
plot_results1(init_fis, final_fis, trn_error, val_error, y_chk, y_hat, i);
|
||||
|
||||
end
|
||||
|
||||
% Show and save table
|
||||
disp(T);
|
||||
writetable(T, 'scenario1_metrics.csv');
|
||||
@@ -0,0 +1,6 @@
|
||||
Metrics,Model_1,Model_2,Model_3,Model_4
|
||||
MSE ,16.9264182248583,17.1406109263442,12.8952585467096,30.4753160187243
|
||||
RMSE,4.11417284819906,4.14012209075339,3.59099687367027,5.520445273592
|
||||
R2 ,0.664677433758077,0.660434146998977,0.744537141243185,0.396265586807705
|
||||
NMSE,0.335322566241923,0.339565853001023,0.255462858756815,0.603734413192295
|
||||
NDEI,0.579070432885261,0.582722792587541,0.505433337599347,0.777003483384917
|
||||
|
@@ -0,0 +1,175 @@
|
||||
%% Scenario2 (TSK - Superconductivity, High-dimensional)
|
||||
%
|
||||
% Assignment 3 in Fuzzy systems
|
||||
%
|
||||
% author:
|
||||
% Christos Choutouridis ΑΕΜ 8997
|
||||
% cchoutou@ece.auth.gr
|
||||
%
|
||||
% Notes:
|
||||
% - 5-fold CV grid-search over (#features, SC radius) with careful
|
||||
% no-leakage pipeline (scaling & ReliefF inside each fold)
|
||||
% - Final training on Train with Validation on Val, Test on Test
|
||||
% - Diagnostics: CV heatmap, error-vs-rules, error-vs-#features
|
||||
% - Classic plots: learning curves, pred vs actual, residuals, MFs (subset)
|
||||
|
||||
clear; clc; close all;
|
||||
|
||||
% Configuration
|
||||
% --------------------------------
|
||||
% Grid of hyperparameters
|
||||
config.feature_grid = [5 8 11 15];
|
||||
config.radius_grid = [0.25 0.5 0.75 1.0];
|
||||
config.K = 5; % 5-fold CV
|
||||
config.Nepochs = 100;
|
||||
|
||||
rng(42,'twister');
|
||||
|
||||
fprintf('Scenario 2 — Superconduct\n');
|
||||
fprintf('===============================\n\n');
|
||||
|
||||
% Load dataset
|
||||
fprintf('Loading dataset (superconduct.csv)...\n');
|
||||
data = load("Datasets/superconduct.csv");
|
||||
fprintf(' Done: %d samples, %d features + 1 target\n\n', size(data,1), size(data,2)-1);
|
||||
|
||||
% Split 60/20/20 and scale using TRAIN stats
|
||||
fprintf('Splitting dataset into Train/Val/Test [0.6/0.2/0.2] ...\n');
|
||||
[X_trn, y_trn, X_val_raw, y_val, X_chk_raw, y_chk] = split_data(data, [0.6 0.2 0.2], 42);
|
||||
fprintf(' Done: Train=%d, Val=%d, Test=%d\n', ...
|
||||
size(X_trn,1), size(X_val_raw,1), size(X_chk_raw,1));
|
||||
|
||||
fprintf('Preprocessing (Min–Max scaling to [0,1]) ...\n');
|
||||
[X_trn_s, X_val_s, X_chk_s, scale_stats] = preprocess_data(X_trn, X_val_raw, X_chk_raw, 1);
|
||||
fprintf(' Done (mode = %s)\n\n', scale_stats.type);
|
||||
|
||||
|
||||
|
||||
fprintf('Starting %d×%d grid-search (%d folds per point)...\n', ...
|
||||
numel(config.feature_grid), numel(config.radius_grid), config.K);
|
||||
|
||||
cv_scores = zeros(numel(config.feature_grid), numel(config.radius_grid));
|
||||
cv_rules = zeros(numel(config.feature_grid), numel(config.radius_grid));
|
||||
|
||||
Ntr = size(X_trn_s,1);
|
||||
folds = cvpartition(Ntr, "KFold", config.K);
|
||||
|
||||
for fi = 1:numel(config.feature_grid)
|
||||
kfeat = config.feature_grid(fi);
|
||||
for rj = 1:numel(config.radius_grid)
|
||||
rad = config.radius_grid(rj);
|
||||
fprintf(' Combo: features=%2d | radius=%.2f ...\n', kfeat, rad);
|
||||
|
||||
fold_err = zeros(config.K,1);
|
||||
fold_rule = zeros(config.K,1);
|
||||
|
||||
for k = 1:config.K
|
||||
fprintf(' Fold %d/%d ... ', k, config.K);
|
||||
|
||||
Itr = training(folds, k);
|
||||
Ivl = test(folds, k);
|
||||
|
||||
% Raw (pre-global-scale) slices
|
||||
x_tr_raw = X_trn(Itr,:); y_tr_f = y_trn(Itr);
|
||||
x_vl_raw = X_trn(Ivl,:); y_vl_f = y_trn(Ivl);
|
||||
|
||||
% Per-fold scaling
|
||||
[x_tr_f, x_vl_f, ~, ~] = preprocess_data(x_tr_raw, x_vl_raw, x_vl_raw, 1);
|
||||
|
||||
% Feature selection (ReliefF)
|
||||
[idxF, ~] = relieff(x_tr_f, y_tr_f, 10);
|
||||
kkeep = min(kfeat, size(x_tr_f,2));
|
||||
sel = idxF(1:kkeep);
|
||||
|
||||
x_tr_f = x_tr_f(:, sel);
|
||||
x_vl_f = x_vl_f(:, sel);
|
||||
|
||||
% Init FIS (SC)
|
||||
gopt = genfisOptions("SubtractiveClustering", ...
|
||||
"ClusterInfluenceRange", rad);
|
||||
init_fis = genfis(x_tr_f, y_tr_f, gopt);
|
||||
|
||||
% Train (Hybrid)
|
||||
aopt = anfisOptions("InitialFis", init_fis, ...
|
||||
"ValidationData", [x_vl_f y_vl_f], ...
|
||||
"EpochNumber", config.Nepochs, ...
|
||||
"OptimizationMethod", 1, ...
|
||||
"DisplayErrorValues", 0, ...
|
||||
"DisplayStepSize", 0);
|
||||
|
||||
[~, ~, ~, vl_fis, vl_err] = anfis([x_tr_f y_tr_f], aopt);
|
||||
|
||||
fold_err(k) = min(vl_err);
|
||||
fold_rule(k) = numel(vl_fis.Rules);
|
||||
fprintf(' Done (val err=%.4g, rules=%d)\n', fold_err(k), fold_rule(k));
|
||||
end
|
||||
|
||||
cv_scores(fi, rj) = mean(fold_err);
|
||||
cv_rules(fi, rj) = round(mean(fold_rule));
|
||||
fprintf(' Mean CV error=%.4g | Mean rules=%d\n\n', ...
|
||||
cv_scores(fi,rj), cv_rules(fi,rj));
|
||||
end
|
||||
end
|
||||
|
||||
fprintf('Grid-search completed.\n');
|
||||
|
||||
% Pick best hyper-parameters
|
||||
[minErr, ix] = min(cv_scores(:));
|
||||
[fi_best, rj_best] = ind2sub(size(cv_scores), ix);
|
||||
best_feats = config.feature_grid(fi_best);
|
||||
best_rad = config.radius_grid(rj_best);
|
||||
fprintf('\n');
|
||||
fprintf(' Best combo found: features=%d, radius=%.2f\n', best_feats, best_rad);
|
||||
fprintf(' Mean CV error : %.4g\n', minErr);
|
||||
fprintf('===========================================\n\n');
|
||||
|
||||
% Final training phase
|
||||
fprintf('Final training using best hyper-parameters...\n');
|
||||
[idx_all, ~] = relieff(X_trn_s, y_trn, 10);
|
||||
kkeep = min(best_feats, size(X_trn_s,2));
|
||||
sel_final = idx_all(1:kkeep);
|
||||
|
||||
XtrF = X_trn_s(:, sel_final);
|
||||
XvlF = X_val_s(:, sel_final);
|
||||
XteF = X_chk_s(:, sel_final);
|
||||
|
||||
fprintf(' Selected %d features (ReliefF top indices)\n', kkeep);
|
||||
fprintf(' Building initial FIS (SC radius=%.2f)...\n', best_rad);
|
||||
|
||||
gopt = genfisOptions("SubtractiveClustering", "ClusterInfluenceRange", best_rad);
|
||||
init_fis = genfis(XtrF, y_trn, gopt);
|
||||
fprintf(' Initial FIS created with %d rules\n', numel(init_fis.Rules));
|
||||
|
||||
aopt = anfisOptions("InitialFis", init_fis, ...
|
||||
"ValidationData", [XvlF y_val], ...
|
||||
"EpochNumber", config.Nepochs, ...
|
||||
"OptimizationMethod", 1, ...
|
||||
"DisplayErrorValues", 1, ...
|
||||
"DisplayStepSize", 0);
|
||||
|
||||
fprintf('Training ANFIS (Hybrid optimization)...\n');
|
||||
[trn_fis, trn_error, ~, fin_fis, val_error] = anfis([XtrF y_trn], aopt);
|
||||
fprintf(' Done. Final FIS has %d rules.\n\n', numel(fin_fis.Rules));
|
||||
|
||||
% Evaluation on TEST
|
||||
fprintf('Evaluating on Test set ...\n');
|
||||
y_hat = evalfis(fin_fis, XteF);
|
||||
[mse, rmse, r2, nmse, ndei] = evaluate(y_hat, y_chk);
|
||||
|
||||
fprintf('\n');
|
||||
fprintf(' FINAL MODEL PERFORMANCE (Test set)\n');
|
||||
fprintf('-------------------------------------------\n');
|
||||
fprintf(' MSE : %g\n', mse);
|
||||
fprintf(' RMSE: %g\n', rmse);
|
||||
fprintf(' R2 : %g\n', r2);
|
||||
fprintf(' NMSE: %g\n', nmse);
|
||||
fprintf(' NDEI: %g\n', ndei);
|
||||
fprintf('\n');
|
||||
|
||||
% Plots
|
||||
fprintf('Generating diagnostic plots ...\n');
|
||||
plot_results2( ...
|
||||
init_fis, fin_fis, trn_error, val_error, y_chk, y_hat, ...
|
||||
config.feature_grid, config.radius_grid, cv_scores, cv_rules, sel_final ...
|
||||
);
|
||||
fprintf(' All plots saved in ./figures_scn2\n');
|
||||
@@ -0,0 +1,25 @@
|
||||
function [X_trn, y_trn, X_val, y_val, X_chk, y_chk] = split_data(data, ratios, seed)
|
||||
%SPLITSET Splits the data-set to train, eval,check data
|
||||
% Randomly split the data set according to ratios
|
||||
%
|
||||
|
||||
% Argument check and default values
|
||||
if nargin < 2 || isempty(ratios), ratios = [0.6 0.2 0.2]; end
|
||||
if nargin == 2, rng(42, 'twister'); end
|
||||
if nargin > 2, rng(seed,'twister'); end
|
||||
|
||||
N = size(data,1); % Scramble the set
|
||||
idx = randperm(N);
|
||||
|
||||
Ntrn = round(ratios(1)*N); % Split positions
|
||||
Nval = round(ratios(2)*N);
|
||||
Itrn = idx(1:Ntrn);
|
||||
Ival = idx(Ntrn+1:Ntrn+Nval);
|
||||
Ichk = idx(Ntrn+Nval+1:end);
|
||||
|
||||
% Pick the data (in X and y)
|
||||
X_trn = data(Itrn, 1:end-1); y_trn = data(Itrn, end);
|
||||
X_val = data(Ival, 1:end-1); y_val = data(Ival, end);
|
||||
X_chk = data(Ichk, 1:end-1); y_chk = data(Ichk, end);
|
||||
|
||||
end
|
||||
Reference in New Issue
Block a user