%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Deep Neural Network: %
% %
% Copyright (C) 2013 Masayuki Tanaka. All rights reserved. %
% mtanaka@ctrl.titech.ac.jp %
% %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clear all;
num = 1000; % num of cases
nodes = [32 16 8 4]; %input to output layers' neuron counts input layer -> output layer
IN = rand(num,32); % 1000 x 32
OUT = rand(num,4); % 1000 x 4
dnn = randDBN( nodes ); % 1 x 1 3 structures of type rbm
%dnn = randDBN( nodes, 'GBDBN' );
nrbm = numel(dnn.rbm); % 3
opts.MaxIter = 20;
opts.BatchSize = num/4; % 250
opts.Verbose = true;
opts.StepRatio = 0.1;
opts.Layer = nrbm-1; % 2
opts.DropOutRate = 0.5;
opts.object = 'CrossEntropy';
dnn = pretrainDBN(dnn, IN, opts);
dnn= SetLinearMapping(dnn, IN, OUT);
opts.Layer = 0;
dnn = trainDBN(dnn, IN, OUT, opts);
rmse = CalcRmse(dnn, IN, OUT);
rmse
%{
debug> size(IN)
ans =
1000 32
debug> size(OUT)
ans =
1000 4
debug> size(dnn)
ans =
1 1
debug> displayentries(dnn)
error: `displayentries' undefined near line 3 column 1
debug> fieldnames(dnn)
ans =
{
[1,1] = type
[2,1] = rbm
}
debug> fieldnames(dnn)
ans =
{
[1,1] = type
[2,1] = rbm
}
debug> size(dnn.type)
ans =
1 5
debug> dnn.type
ans = BBDBN
debug> size(dnn.rbm)
ans =
3 1
debug> dnn.rbm
ans =
{
[1,1] =
scalar structure containing the fields: 32ROWS
type = BBRBM
W =
Columns 1 through 5:
-9.5439e-002 2.2939e-002 7.2628e-002 -1.6725e-001 -1.8585e-001
-6.3338e-002 5.5813e-002 9.9650e-002 -3.2411e-002 1.9054e-001
-2.8721e-002 -3.0517e-001 4.2261e-002 1.8782e-002 1.7918e-002
1.6754e-001 -6.9326e-003 8.4993e-002 1.3912e-001 1.7788e-001
1.2296e-001 2.8321e-003 6.8273e-002 -4.1696e-002 1.8666e-001
-7.1908e-002 7.6576e-002 1.8943e-002 6.0098e-002 -8.2744e-002
-6.1932e-002 -1.9985e-001 1.7706e-002 -1.7833e-001 3.2422e-002
1.4648e-001 1.3261e-001 -8.6583e-002 -5.8321e-002 -7.2068e-002
8.9776e-002 -3.7070e-002 5.6221e-002 -4.0033e-002 9.4303e-003
8.9454e-002 -4.9070e-002 -6.3078e-002 3.1675e-002 -3.0792e-002
1.8290e-001 -1.8830e-001 -8.8328e-002 -1.0021e-002 1.1692e-001
-1.5825e-002 -8.2422e-002 -4.5168e-002 -1.0671e-001 7.1336e-002
-1.4302e-001 1.1588e-002 -1.9893e-002 -1.8817e-002 8.0709e-002
5.8547e-002 1.0674e-001 9.2972e-002 7.3739e-004 -8.8002e-002
-5.2856e-002 5.6408e-002 -2.7927e-002 -5.2166e-002 -2.9046e-002
-2.3111e-002 -2.5470e-002 1.4229e-002 6.3788e-002 -1.0690e-001
1.0172e-001 1.6693e-001 -4.2641e-002 -6.0183e-003 8.3208e-002
9.4523e-003 -1.0553e-001 3.2596e-002 -7.3335e-002 -6.9497e-002
5.5084e-002 -4.9742e-002 6.3967e-002 -3.0585e-002 -2.2389e-003
8.7456e-002 -9.3202e-003 5.7180e-002 -3.3230e-002 -1.3572e-002
1.4025e-001 1.3006e-001 -9.2925e-002 -7.1881e-002 7.9407e-002
4.6632e-002 2.5287e-002 1.8483e-001 -1.2771e-001 4.0131e-002
-4.6402e-002 1.8763e-002 1.2175e-001 3.1542e-002 1.1833e-001
-6.8776e-002 9.5478e-002 1.0184e-001 -1.3870e-001 -1.8715e-002
-1.3226e-002 4.9959e-002 1.3911e-001 -1.2609e-001 -1.1164e-001
5.1048e-003 8.2061e-002 -1.3399e-001 1.4166e-002 4.5473e-002
4.7427e-002 -1.6272e-002 1.2218e-003 2.0601e-002 1.5357e-001
1.2823e-002 9.5251e-002 9.6114e-002 -5.1072e-002 -4.2220e-003
-2.7675e-002 1.3981e-001 -6.3217e-002 1.4558e-002 -1.0971e-001
3.4808e-002 1.0162e-001 -2.7371e-002 -9.5146e-003 -8.1046e-002
-8.9656e-004 5.3675e-002 3.5876e-002 1.1082e-001 -2.6857e-001
-4.4603e-002 5.2974e-002 -1.0659e-001 -8.3114e-002 3.8094e-002
...
Column 16:
1.0590e-001
-5.0035e-002
9.8314e-002
3.4597e-003
1.8471e-001
1.1792e-002
-8.3624e-003
8.8980e-003
8.6437e-002
1.0525e-001
-1.0155e-001
5.0527e-002
-1.0684e-001
8.2844e-002
1.3908e-001
1.8100e-001
-2.6154e-002
1.7376e-002
-1.1795e-001
-1.4004e-001
4.5323e-002
-1.1169e-002
4.8986e-002
7.6310e-002
2.6240e-001
2.4331e-002
7.2168e-003
-1.9634e-002
-9.0908e-002
-1.1385e-002
-1.5389e-001
-8.0523e-002
b =
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
c =
Columns 1 through 18:
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Columns 19 through 32:
0 0 0 0 0 0 0 0 0 0 0 0 0 0
[2,1] =
scalar structure containing the fields: 16ROWS
type = BBRBM
W =
Columns 1 through 5:
2.0102e-001 4.3688e-003 1.2167e-002 3.8655e-002 -1.0856e-001
7.6643e-003 -2.0929e-002 2.1684e-001 2.7952e-002 -1.0454e-001
-1.8838e-001 -1.1319e-001 1.3241e-001 2.4613e-001 -4.1756e-003
-4.1513e-002 -2.4252e-001 4.6049e-002 -7.2507e-002 9.5904e-003
8.3872e-002 9.5719e-002 5.9736e-002 1.0170e-002 -1.9051e-001
-1.6029e-003 -8.5765e-002 1.3034e-001 1.3020e-001 -4.4126e-002
-1.1046e-001 -1.6009e-001 8.4373e-002 1.4629e-001 -4.9012e-002
1.3466e-003 -3.4392e-002 -2.4080e-001 3.8773e-002 -5.2277e-002
8.6146e-004 -2.1853e-002 1.1403e-001 -9.8974e-002 -5.0794e-002
-3.2462e-003 1.0488e-001 3.2717e-002 6.8939e-004 7.9416e-002
-1.1660e-001 -1.1468e-001 1.3624e-001 4.0556e-002 3.1838e-003
4.7486e-002 -1.3841e-001 -5.5895e-002 -1.5473e-001 -4.0310e-002
2.6561e-002 -1.2955e-003 1.6016e-001 6.6688e-002 -1.0138e-002
-1.2804e-001 5.2227e-003 7.8851e-002 -4.8367e-002 2.6254e-004
1.5797e-003 6.0113e-002 3.2811e-002 -1.1808e-001 7.7672e-002
1.4736e-001 -1.7833e-001 -8.0439e-002 -3.6698e-003 -8.5635e-002
Columns 6 through 8:
-4.6001e-002 -6.9266e-002 -2.2738e-001
4.7092e-002 -2.0253e-001 -3.2701e-003
5.1861e-002 -2.1056e-001 -1.8411e-002
-7.1343e-002 1.2337e-001 1.9276e-002
-1.8914e-003 -1.5765e-001 -8.6933e-002
-2.0187e-001 4.3527e-002 -1.7456e-002
1.0685e-001 1.7238e-002 -2.8500e-002
-3.1129e-002 -1.1785e-002 -1.8281e-001
6.6779e-002 -1.9351e-002 1.3626e-001
1.0567e-001 5.0517e-002 -3.2258e-002
-6.2616e-002 -4.1738e-005 -3.2652e-003
2.0329e-003 1.2461e-001 3.1411e-002
8.1571e-002 8.1036e-002 6.0923e-002
5.1120e-002 -4.4483e-002 -1.1280e-001
5.0619e-002 1.1328e-001 1.4899e-001
3.8269e-002 -1.4240e-001 -3.6423e-002
b =
0 0 0 0 0 0 0 0
c =
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
[3,1] =
scalar structure containing the fields: 8ROWS
type = BBRBM
W =
0.0186368 -0.0967863 0.0086856 -0.0792971
-0.0265770 -0.0345573 0.0390910 -0.2094785
0.1285115 0.0039218 0.1239428 -0.0057840
-0.0358651 -0.2717663 -0.0276068 0.0775830
0.0222669 0.0783432 -0.0770633 0.0245969
0.0052533 0.0777309 -0.0423200 -0.0332311
-0.1173345 0.0099952 -0.0984870 -0.1093623
0.0689255 0.0407822 0.0991716 0.0154337
b =
0 0 0 0
c =
0 0 0 0 0 0 0 0
}
debug> nrbm
nrbm = 3
%}
%{
dnn = pretrainDBN(dnn, IN, opts);
***************************************
debug> size(dnn)
ans =
1 1
debug> fieldnames(dnn)
ans =
{
[1,1] = type
[2,1] = rbm
}
debug> size(dnn.type)
ans =
1 5
debug> dnn.type
ans = BBDBN
debug> size(dnn.rbm)
ans =
3 1
debug> dnn.rbm
scalar structure containing the fields:
type = BBRBM
W =
Columns 1 through 5:
-1.4456e-001 1.3457e-001 -1.5983e-001 -2.1078e-001 1.4789e-001
1.9967e-001 3.1571e-002 1.3676e-001 -1.9001e-001 5.2966e-002
-7.9528e-002 -7.3805e-002 1.3651e-001 -1.4727e-002 -5.0989e-002
4.6101e-001 -2.0480e-001 2.2967e-002 -4.4049e-002 -1.4408e-001
-1.5530e-002 5.8309e-002 2.7578e-002 -2.6401e-001 1.6906e-001
-2.3830e-002 7.3771e-002 -4.0720e-002 1.0873e-001 3.2833e-002
-1.1991e-001 -2.7482e-002 -1.7003e-001 -4.4027e-002 -1.6771e-001
-1.4532e-001 1.4782e-001 1.6026e-001 -3.9349e-002 -2.3665e-001
9.6552e-002 9.6118e-002 1.0965e-002 4.8927e-002 1.3123e-001
-1.8393e-001 8.8770e-002 9.7955e-002 1.0479e-001 1.9539e-001
1.5810e-001 8.2610e-003 7.7519e-002 -1.7944e-001 2.0037e-002
-1.8539e-002 2.2513e-002 -1.4977e-001 -6.7766e-002 1.3686e-001
1.3339e-001 -2.9161e-002 7.3629e-002 -7.3675e-002 -1.2211e-001
1.9417e-002 1.4506e-001 -9.1872e-002 -4.9863e-002 -1.4877e-001
6.8912e-002 -7.8553e-002 -1.5891e-001 1.6622e-001 -9.9249e-002
4.6266e-002 1.2187e-001 8.3142e-002 -1.9155e-001 1.5987e-001
1.6179e-002 6.4626e-002 -2.8579e-001 -1.4409e-002 -1.3780e-001
-9.7421e-002 -3.9605e-002 6.2394e-002 -9.7487e-002 2.5618e-002
-2.2313e-002 1.1810e-001 4.9739e-002 -1.9756e-001 -3.7360e-002
1.1228e-001 -2.2252e-001 3.8728e-002 -1.6449e-001 1.3446e-001
-1.0710e-001 2.4637e-001 -3.1594e-002 8.3380e-002 1.0684e-001
-8.0976e-002 -2.2602e-003 1.2758e-001 1.2872e-001 -3.5073e-002
-5.6544e-002 -2.2636e-004 -1.4674e-001 -6.3959e-002 2.7805e-002
3.1768e-001 -6.6169e-002 -7.9164e-002 1.0105e-001 -2.5982e-001
-7.4598e-002 1.4319e-001 2.6001e-001 -1.7961e-001 -1.2858e-001
2.1387e-003 7.2934e-003 -2.0793e-001 -7.9268e-002 2.5656e-001
-1.5909e-001 -1.6167e-002 3.3853e-002 -9.1057e-002 1.8289e-001
3.3792e-002 9.1442e-002 1.0779e-001 1.3711e-001 -3.9855e-002
5.1293e-002 -2.0965e-002 1.6927e-001 -2.1026e-001 -2.0306e-001
-1.4660e-001 1.3455e-002 -5.2939e-002 -3.4808e-002 1.5054e-001
1.0657e-001 -1.4081e-002 -1.0672e-001 1.0797e-001 9.5403e-002
4.2248e-002 -1.3737e-001 -1.6574e-002 1.0592e-001 -1.7172e-001
Columns 6 through 10:
-8.5914e-002 1.0224e-002 2.0198e-001 -1.9529e-001 2.9960e-002
-1.5373e-001 -1.8566e-001 -8.7281e-002 7.6317e-002 -9.4733e-002
-1.3825e-001 6.1988e-002 -1.0458e-001 5.9529e-002 1.6094e-001
2.2550e-002 -1.4377e-001 1.2013e-002 9.3174e-002 -6.1050e-004
-1.1033e-001 3.0341e-001 3.9760e-002 2.7072e-002 -6.9339e-003
2.7696e-001 -8.5717e-002 -3.9012e-002 -1.0869e-002 -4.4833e-002
-9.3750e-003 -1.3472e-001 1.2080e-002 7.9816e-002 -9.2314e-003
1.2384e-001 2.8670e-001 -1.7781e-002 -1.2156e-001 -9.3283e-002
7.7685e-002 -1.8240e-001 2.3156e-002 -1.0981e-001 7.0797e-004
4.6324e-002 1.8556e-002 -2.5338e-001 -1.1749e-001 -1.1827e-001
1.4409e-001 1.0703e-001 -1.9620e-001 -3.1371e-002 -8.1335e-002
-8.4576e-002 8.0624e-002 3.8084e-002 1.2493e-001 9.2895e-003
-1.5890e-001 5.5033e-002 1.2631e-001 1.3307e-001 1.5517e-001
-9.5496e-002 -1.7814e-002 2.5346e-001 1.6847e-001 -5.3799e-002
2.6378e-001 1.8021e-002 -5.2915e-002 -1.3608e-001 2.7665e-002
9.1235e-002 -8.6926e-002 -7.1771e-002 4.5970e-003 9.2774e-002
-9.0323e-002 2.8599e-002 -3.5934e-002 -1.0172e-001 2.7208e-001
9.0449e-002 1.1858e-001 2.3264e-001 -1.5732e-001 -4.1013e-002
-1.1754e-001 1.3658e-001 5.4594e-002 9.1868e-002 -6.6272e-002
-5.5522e-002 -1.4106e-002 -1.0810e-002 -2.2745e-001 1.7566e-002
4.8655e-002 1.3820e-001 -1.2193e-001 -2.5636e-001 -4.7235e-002
9.5951e-002 1.2619e-001 -1.9169e-001 2.1087e-001 -1.9027e-001
3.0819e-002 1.0616e-001 1.0001e-001 1.1986e-001 -3.8117e-001
-1.1276e-001 -5.5931e-002 9.0106e-003 4.9986e-002 -3.6583e-002
1.5102e-001 -2.4015e-001 -5.5786e-002 -2.4883e-002 1.4220e-002
2.2478e-002 -1.2901e-001 -2.4733e-002 -4.6572e-002 7.9630e-002
5.8555e-002 2.7067e-002 -4.7607e-002 4.0240e-002 -4.0228e-002
-1.8334e-001 -1.1726e-001 1.6269e-001 1.0437e-001 -3.3466e-001
-1.3963e-001 -4.8113e-002 -2.4723e-001 1.5928e-001 7.8055e-002
1.4487e-001 -4.0303e-002 3.6406e-002 7.6593e-002 1.0397e-001
-1.0251e-002 -1.1552e-001 3.5737e-002 -1.2316e-001 2.1050e-001
1.9225e-001 -3.8724e-002 -7.8417e-002 5.2468e-002 -9.5619e-002
Columns 11 through 15:
2.4098e-003 -2.6345e-001 -4.6159e-002 1.5794e-001 -3.3164e-003
-1.7913e-001 -2.3133e-002 4.2701e-002 2.3828e-001 1.9139e-001
2.2087e-002 -1.4815e-001 6.2519e-002 -2.5090e-002 8.0639e-003
-2.6046e-001 -1.5779e-001 1.6938e-001 -6.9762e-002 1.4018e-001
8.6579e-002 3.6462e-002 -8.0427e-002 -2.0413e-001 -1.7592e-001
-1.2733e-001 1.0460e-001 -1.7352e-001 -1.4951e-001 -4.0147e-002
2.5901e-002 8.6292e-002 1.3644e-001 2.2090e-001 1.3782e-001
1.1347e-001 4.5161e-002 -4.4923e-002 -3.5492e-002 1.9974e-001
1.7800e-002 3.5112e-002 -9.3266e-002 9.9701e-002 -2.5843e-001
-4.9081e-002 -2.1651e-002 4.9206e-002 4.1672e-002 4.5893e-002
-1.6028e-001 6.9965e-002 -9.9095e-002 -9.8361e-002 1.5742e-002
-2.2755e-001 -1.3005e-001 3.2077e-001 -2.1002e-002 -1.8576e-002
-1.0820e-001 2.2273e-002 1.4682e-001 -1.7895e-001 -2.0020e-001
-1.0212e-001 -9.6141e-002 7.2759e-002 -5.6987e-002 -8.3177e-002
-1.2048e-002 1.4154e-001 1.8797e-001 -7.5862e-002 -2.2842e-001
1.9373e-001 -1.5303e-001 7.4419e-002 -2.0184e-001 -1.2280e-002
8.8329e-002 -5.7753e-002 1.5731e-001 6.5880e-002 -2.9509e-002
5.1196e-002 -6.7306e-002 6.8033e-002 -1.5876e-002 -1.2128e-001
1.3562e-001 1.7572e-001 -9.5858e-002 -5.7359e-002 -4.9138e-002
-1.6606e-002 7.4453e-002 2.2216e-001 -8.0778e-002 9.9390e-002
-1.4711e-001 -2.0306e-002 3.7244e-002 2.0792e-001 -5.3220e-002
-2.0906e-001 -7.5341e-002 3.2026e-002 2.1813e-002 -8.0194e-002
-8.9684e-002 -1.3909e-001 6.9256e-002 1.8700e-001 -3.0837e-002
2.0385e-001 1.0729e-001 5.6024e-002 -1.2941e-001 1.2849e-001
-8.7558e-002 6.1155e-002 2.8669e-002 1.3912e-002 8.1313e-002
-1.1198e-001 1.1722e-001 3.7843e-002 -1.3390e-001 4.9545e-002
1.1977e-001 3.7913e-002 1.4384e-001 -1.8977e-001 -1.2410e-002
-4.7232e-002 -3.8173e-002 5.9145e-002 -3.9736e-002 1.1931e-001
2.6636e-001 1.2498e-001 -2.1388e-002 -2.1876e-001 -4.1982e-003
-2.5778e-002 1.9604e-001 -2.0281e-001 -3.0428e-002 -6.2732e-002
2.0315e-001 -1.7387e-001 1.1643e-002 -9.8135e-002 8.3186e-002
1.8663e-002 2.0766e-002 4.0007e-002 1.4561e-001 1.3943e-002
Column 16:
1.2361e-001
5.3400e-003
3.9879e-002
1.8711e-002
1.3219e-001
-5.1279e-002
1.2698e-001
-5.9815e-002
6.0131e-002
-3.3446e-002
4.0260e-002
-1.0416e-001
-1.4745e-001
-4.7535e-002
-6.4048e-002
-2.2644e-001
-1.8675e-002
6.2748e-002
-2.2745e-001
-7.9144e-002
-1.1377e-001
1.4382e-001
1.2872e-001
-5.1093e-002
-1.5694e-001
5.4393e-002
-1.9857e-001
4.2873e-002
9.1268e-002
5.2551e-002
-1.1336e-001
-1.1882e-001
b =
Columns 1 through 5:
-1.5586e-002 -1.1496e-002 -5.8476e-003 1.5063e-002 -6.7685e-003
Columns 6 through 10:
-5.7987e-003 -6.9171e-003 -7.8332e-003 -1.4338e-004 -4.1969e-003
Columns 11 through 15:
-8.1333e-003 -9.0226e-003 -2.2251e-002 -1.5057e-003 -1.6091e-002
Column 16:
5.6817e-004
c =
Columns 1 through 6:
0.0395916 -0.0773137 0.0312343 -0.0407844 -0.0025245 0.0896613
Columns 7 through 12:
-0.0425101 -0.1091811 -0.0149902 0.0279449 0.0730091 -0.0038612
Columns 13 through 18:
0.0314979 0.0748003 0.0157778 -0.0173519 -0.0313998 -0.0555946
Columns 19 through 24:
-0.0412943 -0.0241210 0.0050137 0.0062377 0.0211264 -0.0494209
Columns 25 through 30:
0.0124015 -0.0334159 0.0270998 0.0090238 0.0583709 -0.0652171
Columns 31 and 32:
-0.0365972 0.0718705
[2,1] =
scalar structure containing the fields:
type = BBRBM
W =
Columns 1 through 6:
-0.1580402 0.1646551 0.1544616 0.0127687 0.0475242 0.0354418
0.1059701 -0.0584153 0.1202782 0.0424994 0.0615194 0.0647433
0.0307732 -0.0471046 -0.2426331 0.1067414 0.0483431 -0.0798042
-0.0731989 -0.0210007 -0.1234566 -0.1587460 0.0332187 -0.2252511
0.0978923 -0.1253813 0.0735168 -0.1567597 -0.0283528 -0.0407749
0.0145115 0.0742178 0.1333530 -0.0073586 0.0637584 0.0700637
0.0969561 0.0043284 -0.1200178 0.1004060 -0.0754928 0.0167230
-0.1466019 0.1610189 -0.1535090 0.0476767 -0.0862423 -0.0467296
-0.0041632 0.1136615 -0.0324736 -0.0667652 0.0516534 -0.0535831
-0.0434391 -0.1836850 0.0426363 0.0200211 0.1795177 -0.1665324
-0.1280892 -0.1473259 0.0141963 -0.0618399 -0.0636552 0.1216574
0.0529018 -0.0140015 0.0391559 -0.0730919 0.2005384 -0.1336324
0.1655463 0.1814396 0.2060112 0.0534279 0.1115840 -0.0068995
-0.2106281 -0.1304211 0.1142619 -0.1817210 -0.1178374 0.1359360
0.0081734 -0.0306936 0.0177958 0.1006536 -0.0477942 -0.0826642
-0.0018466 -0.0195038 -0.0971977 0.0341731 -0.1154291 0.0489353
Columns 7 and 8:
-0.0284126 0.0200409
-0.0930402 0.0764000
0.0341746 -0.0478940
-0.1998030 -0.1460855
0.0696642 0.0428283
0.0259319 -0.0889552
-0.0194931 0.1268362
0.0665831 -0.0574145
0.0629943 -0.0650520
-0.0046916 -0.1639800
-0.0519509 -0.0733316
-0.0539535 -0.1465853
0.2265682 0.0198238
-0.0749331 -0.0618888
-0.0175542 0.0231235
-0.1475020 -0.0175232
b =
Columns 1 through 6:
0.0057857 -0.0016897 0.0027805 -0.0032890 -0.0065730 -0.0010049
Columns 7 and 8:
0.0108362 0.0027294
c =
Columns 1 through 6:
0.0519208 0.1673262 0.1023236 -0.2268309 0.0366060 0.0182441
Columns 7 through 12:
-0.0512072 -0.0443540 0.0091081 -0.1174627 -0.0111318 -0.0157029
Columns 13 through 16:
0.1704334 -0.0983854 -0.0740015 -0.1733870
[3,1] =
scalar structure containing the fields:
type = BBRBM
W =
6.4865e-003 3.1768e-003 -4.8833e-002 -6.2117e-002
-4.3082e-004 -9.2065e-002 -9.6846e-002 3.7618e-002
-7.6594e-002 1.0289e-001 1.0863e-001 3.7683e-002
-6.6581e-002 2.9303e-002 -6.4712e-002 8.4750e-002
1.8016e-001 1.2360e-001 -1.3360e-001 -2.9996e-002
-1.9455e-002 3.0877e-002 -9.0625e-002 -4.8582e-002
-7.6954e-002 -1.1390e-001 4.1767e-002 1.1894e-001
2.4895e-002 -4.6617e-003 -1.1059e-001 -6.8426e-002
b =
-0.0018479 -0.0032890 -0.0012351 -0.0015600
c =
Columns 1 through 6:
0.0308390 0.0916032 0.0644636 -0.0433083 0.0800952 -0.0703054
Columns 7 and 8:
0.0054586 -0.1400794
}
%}
Wednesday, 5 December 2018
Sunday, 25 November 2018
Indexing a matrix with two matrixes
How to index a matrix with matrixes?
A short outline of the matrix indexing will be followed by
a=[1,2,3,4;5,6,7,8;9,10,11,12]
b=[1,2;3,3;2,3]
c=[1;2]
octave:6> a(b,c)
ans =
1 2
9 10
5 6
5 6
9 10
9 10
Indexing Vectors (https://de.mathworks.com/company/newsletters/articles/matrix-indexing-in-matlab.html)
Let's start with the simple case of a vector and a single subscript. The vector is:
v = [16 5 9 4 2 11 7 14];
The subscript can be a single value:
v(3) % Extract the third element
ans =
9
Or the subscript can itself be another vector:
v([1 5 6]) % Extract the first, fifth, and sixth elements
ans =
16 2 11
The crucial point here is item 1 selects item 1 of v which is 16 and
item 5 selects item 5 of v which is 2 and
item 6 selects item 6 of v which is 11.
------------------------------------------------
Let's take an a matrix with octave:
C:\Users\ars>octave -qf
octave:1> a=[1,2,3,4;5,6,7,8;9,10,11,12]
a =
1 2 3 4
5 6 7 8
9 10 11 12
Let's take a b matrix:
octave:2> b=[1,2;3,4;5,6]
b =
1 2
3 4
5 6
Let's take a c matrix:
octave:3> c=[1;2]
c =
1
2
octave:4> a(b,c)
error: A(I,J): row index out of bounds; value 6 out of bound 3
Octave does not like 6 because a has only 3 rows.
Let's take b with only one item bigger than a's row count.
octave:4> b=[1,2;3,4;2,3]
b =
1 2
3 4
2 3
octave:5> a(b,c)
error: A(I,J): row index out of bounds; value 4 out of bound 3
The error repeats. Hence, the items of b matrix must not be
bigger than the row count of a matrix.
Let's take a b matrix with all of its items less or equal to
the row count of a matrix.
octave:5> b=[1,2;3,3;2,3]
b =
1 2
3 3
2 3
Let's try a(b,c) again:
octave:6> a(b,c)
ans =
1 2
9 10
5 6
5 6
9 10
9 10
It works. Let's recall a and c matrixes and analyze this result:
a =
1 2 3 4
5 6 7 8
9 10 11 12
c =
1
2
We have to refer for each of b the related row
and each of c to the realed column of a matrix.
b=1 c=1,2 -> 1 2
b=3 c=1,2 -> 9 10
b=2 c=1,2 -> 5 6
Please notice that b points to each row according to the value of item in sequence
of b 1 5 9 2 6 10 3 7 11 4 8 12
and for each row c proceeds for each column of each row with the value of items in matrix sequence
of c that is 1 2 1 2 1 2 1 2 1 2 ...
namely
b=2 c=1,2 -> 5 6
b=3 c=1,2 -> 9 10
b=3 c=1,2 -> 9 10
Just another example:
octave:7> b=[1,1;2,2;3,3]
b =
1 1
2 2
3 3
Let's recal a and c matrixes:
a =
1 2 3 4
5 6 7 8
9 10 11 12
c =
1
2
octave:8> a(b,c)
ans =
1 2
5 6
9 10
1 2
5 6
9 10
c selects first and second columns. b selects
1 2 3 1 2 3 rows.
Let's increase the rows count of b. And observe:
octave:9> b=[1,1;2,2;3,3;1,1]
b =
1 1
2 2
3 3
1 1
octave:10> a(b,c)
ans =
1 2
5 6
9 10
1 2
1 2
5 6
9 10
1 2
Increasing the size of b (or c) does not cause an
abnormal end. It only increases the selected items
of the output matrix. Infact
the size of the output matrix = size(b) times size(c).
Below given examples show that, the size and shape of the
b and c matrixes a(b,c) does not cause an abend because
the a(b,c) calculation is done according to the item
sequence of b and c matrixes.
octave:11> b=[1,1,1;2,2,2;3,3,3]
b =
1 1 1
2 2 2
3 3 3
octave:12> a(b,c)
ans =
1 2
5 6
9 10
1 2
5 6
9 10
1 2
5 6
9 10
octave:13> c=[1,2,2]
c =
1 2 2
octave:14> a(b,c)
ans =
1 2 2
5 6 6
9 10 10
1 2 2
5 6 6
9 10 10
1 2 2
5 6 6
9 10 10
octave:15> c=[1,2;3,4]
c =
1 2
3 4
octave:16> a(b,c)
ans =
1 3 2 4
5 7 6 8
9 11 10 12
1 3 2 4
5 7 6 8
9 11 10 12
1 3 2 4
5 7 6 8
9 11 10 12
Cheers.
Ali Riza SARAL
arsaral((at))yahoo.com
A short outline of the matrix indexing will be followed by
a=[1,2,3,4;5,6,7,8;9,10,11,12]
b=[1,2;3,3;2,3]
c=[1;2]
octave:6> a(b,c)
ans =
1 2
9 10
5 6
5 6
9 10
9 10
Indexing Vectors (https://de.mathworks.com/company/newsletters/articles/matrix-indexing-in-matlab.html)
Let's start with the simple case of a vector and a single subscript. The vector is:
v = [16 5 9 4 2 11 7 14];
The subscript can be a single value:
v(3) % Extract the third element
ans =
9
Or the subscript can itself be another vector:
v([1 5 6]) % Extract the first, fifth, and sixth elements
ans =
16 2 11
The crucial point here is item 1 selects item 1 of v which is 16 and
item 5 selects item 5 of v which is 2 and
item 6 selects item 6 of v which is 11.
------------------------------------------------
Let's take an a matrix with octave:
C:\Users\ars>octave -qf
octave:1> a=[1,2,3,4;5,6,7,8;9,10,11,12]
a =
1 2 3 4
5 6 7 8
9 10 11 12
Let's take a b matrix:
octave:2> b=[1,2;3,4;5,6]
b =
1 2
3 4
5 6
Let's take a c matrix:
octave:3> c=[1;2]
c =
1
2
octave:4> a(b,c)
error: A(I,J): row index out of bounds; value 6 out of bound 3
Octave does not like 6 because a has only 3 rows.
Let's take b with only one item bigger than a's row count.
octave:4> b=[1,2;3,4;2,3]
b =
1 2
3 4
2 3
octave:5> a(b,c)
error: A(I,J): row index out of bounds; value 4 out of bound 3
The error repeats. Hence, the items of b matrix must not be
bigger than the row count of a matrix.
Let's take a b matrix with all of its items less or equal to
the row count of a matrix.
octave:5> b=[1,2;3,3;2,3]
b =
1 2
3 3
2 3
Let's try a(b,c) again:
octave:6> a(b,c)
ans =
1 2
9 10
5 6
5 6
9 10
9 10
It works. Let's recall a and c matrixes and analyze this result:
a =
1 2 3 4
5 6 7 8
9 10 11 12
c =
1
2
We have to refer for each of b the related row
and each of c to the realed column of a matrix.
b=1 c=1,2 -> 1 2
b=3 c=1,2 -> 9 10
b=2 c=1,2 -> 5 6
Please notice that b points to each row according to the value of item in sequence
of b 1 5 9 2 6 10 3 7 11 4 8 12
and for each row c proceeds for each column of each row with the value of items in matrix sequence
of c that is 1 2 1 2 1 2 1 2 1 2 ...
namely
b=2 c=1,2 -> 5 6
b=3 c=1,2 -> 9 10
b=3 c=1,2 -> 9 10
Just another example:
octave:7> b=[1,1;2,2;3,3]
b =
1 1
2 2
3 3
Let's recal a and c matrixes:
a =
1 2 3 4
5 6 7 8
9 10 11 12
c =
1
2
octave:8> a(b,c)
ans =
1 2
5 6
9 10
1 2
5 6
9 10
c selects first and second columns. b selects
1 2 3 1 2 3 rows.
Let's increase the rows count of b. And observe:
octave:9> b=[1,1;2,2;3,3;1,1]
b =
1 1
2 2
3 3
1 1
octave:10> a(b,c)
ans =
1 2
5 6
9 10
1 2
1 2
5 6
9 10
1 2
Increasing the size of b (or c) does not cause an
abnormal end. It only increases the selected items
of the output matrix. Infact
the size of the output matrix = size(b) times size(c).
Below given examples show that, the size and shape of the
b and c matrixes a(b,c) does not cause an abend because
the a(b,c) calculation is done according to the item
sequence of b and c matrixes.
octave:11> b=[1,1,1;2,2,2;3,3,3]
b =
1 1 1
2 2 2
3 3 3
octave:12> a(b,c)
ans =
1 2
5 6
9 10
1 2
5 6
9 10
1 2
5 6
9 10
octave:13> c=[1,2,2]
c =
1 2 2
octave:14> a(b,c)
ans =
1 2 2
5 6 6
9 10 10
1 2 2
5 6 6
9 10 10
1 2 2
5 6 6
9 10 10
octave:15> c=[1,2;3,4]
c =
1 2
3 4
octave:16> a(b,c)
ans =
1 3 2 4
5 7 6 8
9 11 10 12
1 3 2 4
5 7 6 8
9 11 10 12
1 3 2 4
5 7 6 8
9 11 10 12
Cheers.
Ali Riza SARAL
arsaral((at))yahoo.com
Wednesday, 14 November 2018
Otomatik Cevap Motoru - Cevap Analizi
Mehmet okula gitti.
-------------------
Mehmet nereye gitti?
Mehmet gitti mi?
Okula kim gitti?
Mehmet okula nasıl gitti?
Mehmet ne yaptı?
REF ANALİZİ
====================
0 1 Mehmet İsim (f_önü koşarak %intr. *o refverbalPOS=0 ) \Nesne\ \possible Özne\ <************ÖZNE
1 2 okula İsim->Dolaylı Tümleç -a
2 3 koşarak Zarf -arak -fiilimsi %intr.
3 4 gitti Fiil %dat.intr. *o
3 4 .crlf postPunctuation
Mehmet nereye gitti?
********************
1-Find sentences which has Mehmet as SUBJECT(ÖZNE)
2-Find sentences which has Mehmet and gitti/gitmek(from Morpholojik analysis) as Fiil.
3-Find a Dolaylı Tümleç with -e/a type (nere-ye)
4-Use this tümleç to form an answer sentence -->Mehmet OKULA gitti.
Mehmet gitti mi?
****************
1-Find sentences which has Mehmet as SUBJECT(ÖZNE)
2-Find sentences which has Mehmet and gitti/gitmek(from Morpholojik analysis) as Fiil.
3-If above items are found form answer as YES else NO.
Okula kim gitti?
****************
1-Find sentences which has gitti/gitmek(from Morpholojik analysis) as Fiil.
2-Find sentences which has okula
3-Find Özne in this sentence and ANSWER with the text of that word(s)
Mehmet okula nasıl gitti?
*************************
1-Find sentences which has Mehmet as SUBJECT(ÖZNE)
2-Find sentences which has Mehmet and gitti/gitmek(from Morpholojik analysis) as Fiil.
2-Find sentences which has okula
4- Find Zarf before or after Fiil answer with the text of that word
Mehmet ne yaptı?
****************
1-Find sentences which has Mehmet as SUBJECT(ÖZNE)
2-Limit/warn that the number may be many.
3-Extract Nesne and fiil of these sentences and form sentences such as
'Mehmet gitti'.
-------------------
Mehmet nereye gitti?
Mehmet gitti mi?
Okula kim gitti?
Mehmet okula nasıl gitti?
Mehmet ne yaptı?
REF ANALİZİ
====================
0 1 Mehmet İsim (f_önü koşarak %intr. *o refverbalPOS=0 ) \Nesne\ \possible Özne\ <************ÖZNE
1 2 okula İsim->Dolaylı Tümleç -a
2 3 koşarak Zarf -arak -fiilimsi %intr.
3 4 gitti Fiil %dat.intr. *o
3 4 .crlf postPunctuation
Mehmet nereye gitti?
********************
1-Find sentences which has Mehmet as SUBJECT(ÖZNE)
2-Find sentences which has Mehmet and gitti/gitmek(from Morpholojik analysis) as Fiil.
3-Find a Dolaylı Tümleç with -e/a type (nere-ye)
4-Use this tümleç to form an answer sentence -->Mehmet OKULA gitti.
Mehmet gitti mi?
****************
1-Find sentences which has Mehmet as SUBJECT(ÖZNE)
2-Find sentences which has Mehmet and gitti/gitmek(from Morpholojik analysis) as Fiil.
3-If above items are found form answer as YES else NO.
Okula kim gitti?
****************
1-Find sentences which has gitti/gitmek(from Morpholojik analysis) as Fiil.
2-Find sentences which has okula
3-Find Özne in this sentence and ANSWER with the text of that word(s)
Mehmet okula nasıl gitti?
*************************
1-Find sentences which has Mehmet as SUBJECT(ÖZNE)
2-Find sentences which has Mehmet and gitti/gitmek(from Morpholojik analysis) as Fiil.
2-Find sentences which has okula
4- Find Zarf before or after Fiil answer with the text of that word
Mehmet ne yaptı?
****************
1-Find sentences which has Mehmet as SUBJECT(ÖZNE)
2-Limit/warn that the number may be many.
3-Extract Nesne and fiil of these sentences and form sentences such as
'Mehmet gitti'.
Otomatik Cevap Motoru - Soru Dosyası Analizi
REFERANS METİN
==============
Mehmet okula koşarak gitti.
Ayşe topu bana attı.
O matematik dersini çok çalışır.
Ahmet heyecanlı bir çocuktur.
Mehmet okula gitti.
-------------------
Mehmet nereye gitti?
Mehmet gitti mi?
Okula kim gitti?
Mehmet okula nasıl gitti?
Mehmet ne yaptı?
Ayşe topu bana attı.
--------------------
Kim topu attı?
Kim topu bana attı?
Bana topu kim attı?
Ayşe bana ne attı?
Ayşe ne yaptı?
Top ne oldu?
O matematik dersini çok çalışır.
-------------------------------
Matematik dersini kim çok çalışır?
O neyi çok çalışır?
O hangi dersi çok çalışır?
O matematik dersini ne yapar?
O ne yapar?
Ahmet heyecanlı bir çocuktur.
-----------------------------
Ahmet nasıl bir çocuktur?
Ahmet kimdir?
==============
Mehmet okula koşarak gitti.
Ayşe topu bana attı.
O matematik dersini çok çalışır.
Ahmet heyecanlı bir çocuktur.
Mehmet okula gitti.
-------------------
Mehmet nereye gitti?
Mehmet gitti mi?
Okula kim gitti?
Mehmet okula nasıl gitti?
Mehmet ne yaptı?
Ayşe topu bana attı.
--------------------
Kim topu attı?
Kim topu bana attı?
Bana topu kim attı?
Ayşe bana ne attı?
Ayşe ne yaptı?
Top ne oldu?
O matematik dersini çok çalışır.
-------------------------------
Matematik dersini kim çok çalışır?
O neyi çok çalışır?
O hangi dersi çok çalışır?
O matematik dersini ne yapar?
O ne yapar?
Ahmet heyecanlı bir çocuktur.
-----------------------------
Ahmet nasıl bir çocuktur?
Ahmet kimdir?
Otomatik Cevap Motoru - Ref Dosyası Analizi
sentence= 0<<****************************dumpSentenceAsText
Mehmet okula koşarak gitti.
sentence= 0<<----------------------------dumpstructlist p="">0 1 Mehmet İsim (f_önü koşarak %intr. *o refverbalPOS=0 ) \Nesne\ \possible Özne\ <************ÖZNE
1 2 okula İsim->Dolaylı Tümleç -a
2 3 koşarak Zarf -arak -fiilimsi %intr.
3 4 gitti Fiil %dat.intr. *o
3 4 .crlf postPunctuation
sentence= 1<<****************************dumpSentenceAsText
Ayşe topu bana attı.
sentence= 1<<----------------------------dumpstructlist p="">0 2 Ayşe topu İsim Tamlaması -i/ı/u/ü (f_önü attı %tr.intr.dat. *o ref ) \Nesne\ <************NESNE
2 3 bana Zamir->Dolaylı Tümleç -a
3 4 attı Fiil %tr.intr.dat. *o
3 4 .crlf postPunctuation
sentence= 2<<****************************dumpSentenceAsText
O matematik dersini çok çalışır.
sentence= 2<<----------------------------dumpstructlist p="">0 1 O Zamir \possible Özne\ ( çalışır %tr.intr. *o ref )<************ÖZNE
1 3 matematik dersini Sıfat Tamlaması->Nesne -ini (f_önü çalışır %tr.intr. *o ref ) \Nesne\ <************NESNE
3 4 çok Zarf
4 5 çalışır Fiil %tr.intr. *o
4 5 .crlf postPunctuation
sentence= 3<<****************************dumpSentenceAsText
Ahmet heyecanlı bir çocuktur.
sentence= 3<<----------------------------dumpstructlist p="">0 1 Ahmet İsim (f_önü heyecanlı bir çocuktur %intr. *o ref ) \Nesne\ \possible Özne\ <************ÖZNE
1 4 heyecanlı bir çocuktur Fiil İsim/Sıfat Tamlaması Türevi %intr. -tur
3 4 .crlfcrlfcrlfcrlfcrlf postPunctuation
BUILD SUCCESSFUL (total time: 1 second)
MORPHOLOJIK(kelime yapısı) VERİLERİ
===================================
sentence= 0<<****************************dumpSentenceAsText
Mehmet okula koşarak gitti.
sentence= 0<<----------------------------morphologic analysis="" p="">0 0 0 Mehmet
R is. ===>Mehmet mehmet
1 0 1 okula
N is. ===>okula =okul -a ^Noun->Noun
2 0 2 koşarak
V f. ===>koşarak =koş -arak ^Verb->Adverb ->Adverb %intr.
3 0 3 gitti.
V f. ===>gitti =git -ti ^DiliGeçmiş %dat.intr.
sentence= 1<<****************************dumpSentenceAsText
Ayşe topu bana attı.
sentence= 1<<----------------------------morphologic analysis="" p="">4 1 0 Ayşe
Ayşe-->ayşe
R is. ===>Ayşe ayşe
5 1 1 topu
R is. ===>topu top -u
6 1 2 bana
R zm. ===>bana ben -a
7 1 3 attı.
V f. ===>attı =at -tı ^DiliGeçmiş %tr.intr.dat.
sentence= 2<<****************************dumpSentenceAsText
O matematik dersini çok çalışır.
sentence= 2<<----------------------------morphologic analysis="" p="">8 2 0 O
O-->o
R zm. ===>O o
9 2 1 matematik
R sf. ===>matematik
10 2 2 dersini
N is. ===>dersini =ders -ini ^Noun->Noun
11 2 3 çok
R zf. sf. ===>çok
12 2 4 çalışır.
V f. ===>çalışır =çalış -ır ^IRliGenişZ %tr.intr.
sentence= 3<<****************************dumpSentenceAsText
Ahmet heyecanlı bir çocuktur.
sentence= 3<<----------------------------morphologic analysis="" p="">13 3 0 Ahmet
Ahmet-->ahmet
R is. ===>Ahmet ahmet
14 3 1 heyecanlı
R sf. ===>heyecanlı
N is. ===>heyecanlı =heyecan -lı ^Noun->Noun/Adjective
15 3 2 bir
R sf. ===>bir
16 3 3 çocuktur.
N is. ===>çocuktur =çocuk -tur ^Noun->Verb %intr.
Mehmet okula koşarak gitti.
sentence= 0<<----------------------------dumpstructlist p="">0 1 Mehmet İsim (f_önü koşarak %intr. *o refverbalPOS=0 ) \Nesne\ \possible Özne\ <************ÖZNE
1 2 okula İsim->Dolaylı Tümleç -a
2 3 koşarak Zarf -arak -fiilimsi %intr.
3 4 gitti Fiil %dat.intr. *o
3 4 .crlf postPunctuation
sentence= 1<<****************************dumpSentenceAsText
Ayşe topu bana attı.
sentence= 1<<----------------------------dumpstructlist p="">0 2 Ayşe topu İsim Tamlaması -i/ı/u/ü (f_önü attı %tr.intr.dat. *o ref ) \Nesne\ <************NESNE
2 3 bana Zamir->Dolaylı Tümleç -a
3 4 attı Fiil %tr.intr.dat. *o
3 4 .crlf postPunctuation
sentence= 2<<****************************dumpSentenceAsText
O matematik dersini çok çalışır.
sentence= 2<<----------------------------dumpstructlist p="">0 1 O Zamir \possible Özne\ ( çalışır %tr.intr. *o ref )<************ÖZNE
1 3 matematik dersini Sıfat Tamlaması->Nesne -ini (f_önü çalışır %tr.intr. *o ref ) \Nesne\ <************NESNE
3 4 çok Zarf
4 5 çalışır Fiil %tr.intr. *o
4 5 .crlf postPunctuation
sentence= 3<<****************************dumpSentenceAsText
Ahmet heyecanlı bir çocuktur.
sentence= 3<<----------------------------dumpstructlist p="">0 1 Ahmet İsim (f_önü heyecanlı bir çocuktur %intr. *o ref ) \Nesne\ \possible Özne\ <************ÖZNE
1 4 heyecanlı bir çocuktur Fiil İsim/Sıfat Tamlaması Türevi %intr. -tur
3 4 .crlfcrlfcrlfcrlfcrlf postPunctuation
BUILD SUCCESSFUL (total time: 1 second)
MORPHOLOJIK(kelime yapısı) VERİLERİ
===================================
sentence= 0<<****************************dumpSentenceAsText
Mehmet okula koşarak gitti.
sentence= 0<<----------------------------morphologic analysis="" p="">0 0 0 Mehmet
R is. ===>Mehmet mehmet
1 0 1 okula
N is. ===>okula =okul -a ^Noun->Noun
2 0 2 koşarak
V f. ===>koşarak =koş -arak ^Verb->Adverb ->Adverb %intr.
3 0 3 gitti.
V f. ===>gitti =git -ti ^DiliGeçmiş %dat.intr.
sentence= 1<<****************************dumpSentenceAsText
Ayşe topu bana attı.
sentence= 1<<----------------------------morphologic analysis="" p="">4 1 0 Ayşe
Ayşe-->ayşe
R is. ===>Ayşe ayşe
5 1 1 topu
R is. ===>topu top -u
6 1 2 bana
R zm. ===>bana ben -a
7 1 3 attı.
V f. ===>attı =at -tı ^DiliGeçmiş %tr.intr.dat.
sentence= 2<<****************************dumpSentenceAsText
O matematik dersini çok çalışır.
sentence= 2<<----------------------------morphologic analysis="" p="">8 2 0 O
O-->o
R zm. ===>O o
9 2 1 matematik
R sf. ===>matematik
10 2 2 dersini
N is. ===>dersini =ders -ini ^Noun->Noun
11 2 3 çok
R zf. sf. ===>çok
12 2 4 çalışır.
V f. ===>çalışır =çalış -ır ^IRliGenişZ %tr.intr.
sentence= 3<<****************************dumpSentenceAsText
Ahmet heyecanlı bir çocuktur.
sentence= 3<<----------------------------morphologic analysis="" p="">13 3 0 Ahmet
Ahmet-->ahmet
R is. ===>Ahmet ahmet
14 3 1 heyecanlı
R sf. ===>heyecanlı
N is. ===>heyecanlı =heyecan -lı ^Noun->Noun/Adjective
15 3 2 bir
R sf. ===>bir
16 3 3 çocuktur.
N is. ===>çocuktur =çocuk -tur ^Noun->Verb %intr.
Otomatik Cevap Motoru - Ref Dosyası
Mehmet okula koşarak gitti.
Ayşe topu bana attı.
O matematik dersini çok çalışır.
Ahmet heyecanlı bir çocuktur.
Ayşe topu bana attı.
O matematik dersini çok çalışır.
Ahmet heyecanlı bir çocuktur.
Wednesday, 7 November 2018
mnistclassify analysis 4
CG_CLASSIFY.m
% Version 1.000
%
% Code provided by Ruslan Salakhutdinov and Geoff Hinton
%
% Permission is granted for anyone to copy, use, modify, or distribute this
% program and accompanying programs and documents for any purpose, provided
% this copyright notice is retained and prominently displayed, along with
% a note saying that the original programs are available from our
% web page.
% The programs and documents are distributed without any warranty, express or
% implied. As the programs were written for research purposes only, they have
% not been tested to the degree that would be advisable in any important
% application. All use of these programs is entirely at the user's own risk.
function [f, df] = CG_CLASSIFY(VV,Dim,XX,target);
l1 = Dim(1); % 784
l2 = Dim(2); % 250
l3= Dim(3); % 250
l4= Dim(4); 5 50
l5= Dim(5); % 10
N = size(XX,1); % 1000 785
% Do decomversion.
w1 = reshape(VV(1:(l1+1)*l2),l1+1,l2); % 785 250 = 272060(1:785*250, 785,250)
xxx = (l1+1)*l2; % 785 * 250
w2 = reshape(VV(xxx+1:xxx+(l2+1)*l3),l2+1,l3); % 251 250
xxx = xxx+(l2+1)*l3;
w3 = reshape(VV(xxx+1:xxx+(l3+1)*l4),l3+1,l4); % 251 50
xxx = xxx+(l3+1)*l4;
w_class = reshape(VV(xxx+1:xxx+(l4+1)*l5),l4+1,l5); % 51 10
XX = [XX ones(N,1)]; %1000 785 data+1col
w1probs = 1./(1 + exp(-XX*w1)); w1probs = [w1probs ones(N,1)]; % 1000 785 * 785 250 -> 1000 251
w2probs = 1./(1 + exp(-w1probs*w2)); w2probs = [w2probs ones(N,1)]; % 1000 251 * 251 250 -> 1000 251
w3probs = 1./(1 + exp(-w2probs*w3)); w3probs = [w3probs ones(N,1)]; % 1000 251 * 251 50 -> 1000 51
targetout = exp(w3probs*w_class); % 1000 51 * 51 10 = 1000 10
targetout = targetout./repmat(sum(targetout,2),1,10); % repmat((1000 1),1,10) normalize
f = -sum(sum( target(:,1:end).*log(targetout))) ; % cross enthropy
IO = (targetout-target(:,1:end)); % 1000 10
Ix_class=IO;
dw_class = w3probs'*Ix_class; % 1000 51' * 1000 10 = 51 10
Ix3 = (Ix_class*w_class').*w3probs.*(1-w3probs); % 1000 10 * 51 10' = 1000 51 .* 1000 51 = 1000 51
Ix3 = Ix3(:,1:end-1); % 1000 50
dw3 = w2probs'*Ix3; %1000 251' * 1000 50 = 251 50
Ix2 = (Ix3*w3').*w2probs.*(1-w2probs); % 1000 50 * 251 50' .* 1000 251 = 1000 251
Ix2 = Ix2(:,1:end-1); % 1000 250
dw2 = w1probs'*Ix2; % 1000 251' * 1000 250 = 251 250
Ix1 = (Ix2*w2').*w1probs.*(1-w1probs); % 1000 250 * 251 250' .* 1000 251 = 1000 251
Ix1 = Ix1(:,1:end-1); % 1000 250
dw1 = XX'*Ix1; % 1000 785' * 1000 250 = 785 250
df = [dw1(:)' dw2(:)' dw3(:)' dw_class(:)']'; % 272060 1
% Version 1.000
%
% Code provided by Ruslan Salakhutdinov and Geoff Hinton
%
% Permission is granted for anyone to copy, use, modify, or distribute this
% program and accompanying programs and documents for any purpose, provided
% this copyright notice is retained and prominently displayed, along with
% a note saying that the original programs are available from our
% web page.
% The programs and documents are distributed without any warranty, express or
% implied. As the programs were written for research purposes only, they have
% not been tested to the degree that would be advisable in any important
% application. All use of these programs is entirely at the user's own risk.
function [f, df] = CG_CLASSIFY(VV,Dim,XX,target);
l1 = Dim(1); % 784
l2 = Dim(2); % 250
l3= Dim(3); % 250
l4= Dim(4); 5 50
l5= Dim(5); % 10
N = size(XX,1); % 1000 785
% Do decomversion.
w1 = reshape(VV(1:(l1+1)*l2),l1+1,l2); % 785 250 = 272060(1:785*250, 785,250)
xxx = (l1+1)*l2; % 785 * 250
w2 = reshape(VV(xxx+1:xxx+(l2+1)*l3),l2+1,l3); % 251 250
xxx = xxx+(l2+1)*l3;
w3 = reshape(VV(xxx+1:xxx+(l3+1)*l4),l3+1,l4); % 251 50
xxx = xxx+(l3+1)*l4;
w_class = reshape(VV(xxx+1:xxx+(l4+1)*l5),l4+1,l5); % 51 10
XX = [XX ones(N,1)]; %1000 785 data+1col
w1probs = 1./(1 + exp(-XX*w1)); w1probs = [w1probs ones(N,1)]; % 1000 785 * 785 250 -> 1000 251
w2probs = 1./(1 + exp(-w1probs*w2)); w2probs = [w2probs ones(N,1)]; % 1000 251 * 251 250 -> 1000 251
w3probs = 1./(1 + exp(-w2probs*w3)); w3probs = [w3probs ones(N,1)]; % 1000 251 * 251 50 -> 1000 51
targetout = exp(w3probs*w_class); % 1000 51 * 51 10 = 1000 10
targetout = targetout./repmat(sum(targetout,2),1,10); % repmat((1000 1),1,10) normalize
f = -sum(sum( target(:,1:end).*log(targetout))) ; % cross enthropy
IO = (targetout-target(:,1:end)); % 1000 10
Ix_class=IO;
dw_class = w3probs'*Ix_class; % 1000 51' * 1000 10 = 51 10
Ix3 = (Ix_class*w_class').*w3probs.*(1-w3probs); % 1000 10 * 51 10' = 1000 51 .* 1000 51 = 1000 51
Ix3 = Ix3(:,1:end-1); % 1000 50
dw3 = w2probs'*Ix3; %1000 251' * 1000 50 = 251 50
Ix2 = (Ix3*w3').*w2probs.*(1-w2probs); % 1000 50 * 251 50' .* 1000 251 = 1000 251
Ix2 = Ix2(:,1:end-1); % 1000 250
dw2 = w1probs'*Ix2; % 1000 251' * 1000 250 = 251 250
Ix1 = (Ix2*w2').*w1probs.*(1-w1probs); % 1000 250 * 251 250' .* 1000 251 = 1000 251
Ix1 = Ix1(:,1:end-1); % 1000 250
dw1 = XX'*Ix1; % 1000 785' * 1000 250 = 785 250
df = [dw1(:)' dw2(:)' dw3(:)' dw_class(:)']'; % 272060 1
mnistclassify analysis 3
CG_CLASSIFY_INIT.m
% Version 1.000
%
% Code provided by Ruslan Salakhutdinov and Geoff Hinton
%
% Permission is granted for anyone to copy, use, modify, or distribute this
% program and accompanying programs and documents for any purpose, provided
% this copyright notice is retained and prominently displayed, along with
% a note saying that the original programs are available from our
% web page.
% The programs and documents are distributed without any warranty, express or
% implied. As the programs were written for research purposes only, they have
% not been tested to the degree that would be advisable in any important
% application. All use of these programs is entirely at the user's own risk.
function [f, df] = CG_CLASSIFY_INIT(VV,Dim,w3probs,target); % 510 1, 2 1, 1000 51, 1000 10
l1 = Dim(1); % 50
l2 = Dim(2); % 10
N = size(w3probs,1); % 1000
% Do decomversion.
w_class = reshape(VV,l1+1,l2); % 51 10
w3probs = [w3probs ones(N,1)]; % 1000 51
targetout = exp(w3probs*w_class); % 1000 51 * 51 10 = 1000 10
targetout = targetout./repmat(sum(targetout,2),1,10); % repmat((1000 1),1,10) normalize
%
f = -sum(sum( target(:,1:end).*log(targetout))) ; % cross enthropy
IO = (targetout-target(:,1:end)); % 1000 10
Ix_class=IO;
dw_class = w3probs'*Ix_class; % 1000 51' * 1000 10 = 51 10
df = [dw_class(:)']'; % 510 1
%{
debug> a=[1,2,3;4,5,6]
a =
1 2 3
4 5 6
debug> sum(a,2)
ans =
6
15
debug> repmat(sum(a,2),1,10)
ans =
6 6 6 6 6 6 6 6 6 6
15 15 15 15 15 15 15 15 15 15
%}
% Version 1.000
%
% Code provided by Ruslan Salakhutdinov and Geoff Hinton
%
% Permission is granted for anyone to copy, use, modify, or distribute this
% program and accompanying programs and documents for any purpose, provided
% this copyright notice is retained and prominently displayed, along with
% a note saying that the original programs are available from our
% web page.
% The programs and documents are distributed without any warranty, express or
% implied. As the programs were written for research purposes only, they have
% not been tested to the degree that would be advisable in any important
% application. All use of these programs is entirely at the user's own risk.
function [f, df] = CG_CLASSIFY_INIT(VV,Dim,w3probs,target); % 510 1, 2 1, 1000 51, 1000 10
l1 = Dim(1); % 50
l2 = Dim(2); % 10
N = size(w3probs,1); % 1000
% Do decomversion.
w_class = reshape(VV,l1+1,l2); % 51 10
w3probs = [w3probs ones(N,1)]; % 1000 51
targetout = exp(w3probs*w_class); % 1000 51 * 51 10 = 1000 10
targetout = targetout./repmat(sum(targetout,2),1,10); % repmat((1000 1),1,10) normalize
%
f = -sum(sum( target(:,1:end).*log(targetout))) ; % cross enthropy
IO = (targetout-target(:,1:end)); % 1000 10
Ix_class=IO;
dw_class = w3probs'*Ix_class; % 1000 51' * 1000 10 = 51 10
df = [dw_class(:)']'; % 510 1
%{
debug> a=[1,2,3;4,5,6]
a =
1 2 3
4 5 6
debug> sum(a,2)
ans =
6
15
debug> repmat(sum(a,2),1,10)
ans =
6 6 6 6 6 6 6 6 6 6
15 15 15 15 15 15 15 15 15 15
%}
mnistclassify analysis 2
backpropclassify.m
% Version 1.000
%
% Code provided by Ruslan Salakhutdinov and Geoff Hinton
%
% Permission is granted for anyone to copy, use, modify, or distribute this
% program and accompanying programs and documents for any purpose, provided
% this copyright notice is retained and prominently displayed, along with
% a note saying that the original programs are available from our
% web page.
% The programs and documents are distributed without any warranty, express or
% implied. As the programs were written for research purposes only, they have
% not been tested to the degree that would be advisable in any important
% application. All use of these programs is entirely at the user's own risk.
% This program fine-tunes an autoencoder with backpropagation.
% Weights of the autoencoder are going to be saved in mnist_weights.mat
% and trainig and test reconstruction errors in mnist_error.mat
% You can also set maxepoch, default value is 200 as in our paper.
maxepoch=2; %maxepoch=200;
fprintf(1,'\nTraining discriminative model on MNIST by minimizing cross entropy error. \n');
fprintf(1,'60 batches of 1000 cases each. \n');
load mnistvhclassify
load mnisthpclassify
load mnisthp2classify
makebatches;
[numcases numdims numbatches]=size(batchdata); % 100 784 600
N=numcases; % 100
%%%% PREINITIALIZE WEIGHTS OF THE DISCRIMINATIVE MODEL%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
w1=[vishid; hidrecbiases]; % 784 250 + 1 250 = 785 250
w2=[hidpen; penrecbiases]; % 250 250 + 1 250 = 251 250
w3=[hidpen2; penrecbiases2]; % 250 500 + 1 500 = 251 500
w_class = 0.1*randn(size(w3,2)+1,10); % randn(501,10) = 501 10
%%%%%%%%%% END OF PREINITIALIZATION OF WEIGHTS %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
l1=size(w1,1)-1; % 784
l2=size(w2,1)-1; % 250
l3=size(w3,1)-1; % 250
l4=size(w_class,1)-1; % 500
l5=10;
test_err=[];
train_err=[];
for epoch = 1:maxepoch
%%%%%%%%%%%%%%%%%%%% COMPUTE TRAINING MISCLASSIFICATION ERROR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
err=0;
err_cr=0;
counter=0;
[numcases numdims numbatches]=size(batchdata); % 100 784 600
N=numcases; % 100
for batch = 1:numbatches % 1 : 600
data = [batchdata(:,:,batch)]; % 100 784
target = [batchtargets(:,:,batch)]; % 100 10 600
data = [data ones(N,1)]; % 100 785
w1probs = 1./(1 + exp(-data*w1)); w1probs = [w1probs ones(N,1)]; % 100 785 * 785 250 += 100 251
w2probs = 1./(1 + exp(-w1probs*w2)); w2probs = [w2probs ones(N,1)]; % 100 251 * 251 250 += 100 251
w3probs = 1./(1 + exp(-w2probs*w3)); w3probs = [w3probs ones(N,1)]; % 100 251 * 251 500 += 100 501
targetout = exp(w3probs*w_class); % 100 501 * 501 10 = 100 10
targetout = targetout./repmat(sum(targetout,2),1,10); % 100 10 = 100 10 ./ repmat ( 100 1), 1 10) = 100 10
[I J]=max(targetout,[],2); % 100 1 , 100 1 = 100 1 -->I has the value J has the sequence
[I1 J1]=max(target,[],2); % max(100 10,[],2) 100 1
counter=counter+length(find(J==J1)); % =6 for the first batch
err_cr = err_cr- sum(sum( target(:,1:end).*log(targetout))) ; %cross entrophy
end
train_err(epoch)=(numcases*numbatches-counter); % total number of errors for all the batches in this epoche
train_crerr(epoch)=err_cr/numbatches; % total cross enthropy error for the complete batchdata in this epoche
%%%%%%%%%%%%%% END OF COMPUTING TRAINING MISCLASSIFICATION ERROR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%% COMPUTE TEST MISCLASSIFICATION ERROR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
err=0;
err_cr=0;
counter=0;
[testnumcases testnumdims testnumbatches]=size(testbatchdata); % 100 784 100
N=testnumcases; % 100
for batch = 1:testnumbatches % 1: 100
data = [testbatchdata(:,:,batch)]; % 100 784
target = [testbatchtargets(:,:,batch)]; % 100 10 = (100 10 100(:,:,batch)
data = [data ones(N,1)]; % 100 785
w1probs = 1./(1 + exp(-data*w1)); w1probs = [w1probs ones(N,1)]; % 100 785 *785 250 = 100 250 -> 100 251
w2probs = 1./(1 + exp(-w1probs*w2)); w2probs = [w2probs ones(N,1)]; % 100 251 * 251 250 = 100 250 -> 100 251
w3probs = 1./(1 + exp(-w2probs*w3)); w3probs = [w3probs ones(N,1)]; % 100 251 * 251 50 = 100 50 -> 100 51
targetout = exp(w3probs*w_class); % 100 51 * 51 10 = 100 10
targetout = targetout./repmat(sum(targetout,2),1,10); % = 100 10 ./ repmat ( 100 1), 1 10) = 100 10
[I J]=max(targetout,[],2); % 100 1 , 100 1 = 100 1 -->I has the value J has the sequence
[I1 J1]=max(target,[],2); % max(100 10,[],2) 100 1
counter=counter+length(find(J==J1)); % =9 for the first batch
err_cr = err_cr- sum(sum( target(:,1:end).*log(targetout))); %cross entrophy
end
test_err(epoch)=(testnumcases*testnumbatches-counter); % total number of errors for all the batches in this epoche
test_crerr(epoch)=err_cr/testnumbatches; % total cross enthropy error for the complete batchdata in this epoche
fprintf(1,'Before epoch %d Train # misclassified: %d (from %d). Test # misclassified: %d (from %d) \t \t \n',...
epoch,train_err(epoch),numcases*numbatches,test_err(epoch),testnumcases*testnumbatches);
%%%%%%%%%%%%%% END OF COMPUTING TEST MISCLASSIFICATION ERROR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
tt=0;
for batch = 1:numbatches/10
fprintf(1,'epoch %d batch %d\n',epoch,batch);
%%%%%%%%%%% COMBINE 10 MINIBATCHES INTO 1 LARGER MINIBATCH %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
tt=tt+1;
data=[];
targets=[];
for kk=1:10
data=[data
batchdata(:,:,(tt-1)*10+kk)]; % 1000 784
targets=[targets
batchtargets(:,:,(tt-1)*10+kk)]; % 1000 10
end
%%%%%%%%%%%%%%% PERFORM CONJUGATE GRADIENT WITH 3 LINESEARCHES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
max_iter=3;
if epoch<2 6="" fixed.="" holding="" irst="" nbsp="" original="" other="" p="" top-level="" update="" weights=""> N = size(data,1); % 1000 /1000 784)
XX = [data ones(N,1)]; % 1000 785
w1probs = 1./(1 + exp(-XX*w1)); w1probs = [w1probs ones(N,1)]; % 1000 785 * 785 250 = 1000 250 -> 1000 251
w2probs = 1./(1 + exp(-w1probs*w2)); w2probs = [w2probs ones(N,1)]; % 1000 251 * 251 250 = 1000 250 -> 1000 251
w3probs = 1./(1 + exp(-w2probs*w3)); %w3probs = [w3probs ones(N,1)]; % 1000 251 * 251 50 = 1000 50 -> 1000 51
VV = [w_class(:)']'; % 51 10 = 510 1
Dim = [l4; l5]; % 2 1
[X, fX] = minimize(VV,'CG_CLASSIFY_INIT',max_iter,Dim,w3probs,targets); % 510 1, 4 1 = min(510 1,'CG_CLASS...',3, 2 1, 1000 51, 1000 10
w_class = reshape(X,l4+1,l5); %reshape(X,51,10)= 51 10
else
VV = [w1(:)' w2(:)' w3(:)' w_class(:)']'; % 272060 1
Dim = [l1; l2; l3; l4; l5]; % 5 1
[X, fX] = minimize(VV,'CG_CLASSIFY',max_iter,Dim,data,targets); %[% 272060 1, 4 1]=mini..(272060 1,'CG_CLA..', 3, 5 1, 1000 784, 1000 10);
w1 = reshape(X(1:(l1+1)*l2),l1+1,l2); %reshape(272060(1 : 785*250,785,250) = 785 250
xxx = (l1+1)*l2; % 785 * 250
w2 = reshape(X(xxx+1:xxx+(l2+1)*l3),l2+1,l3); % reshape(X(251 : 251 * 250), 251, 250); = 251 250
xxx = xxx+(l2+1)*l3; % 785 * 250 + 251 * 250
w3 = reshape(X(xxx+1:xxx+(l3+1)*l4),l3+1,l4); % reshape(X(xxx+1: xxx+251*50, 251,50) = 251, 50
xxx = xxx+(l3+1)*l4; % 271550
w_class = reshape(X(xxx+1:xxx+(l4+1)*l5),l4+1,l5); % 51 10
end
%%%%%%%%%%%%%%% END OF CONJUGATE GRADIENT WITH 3 LINESEARCHES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
end
save mnistclassify_weights w1 w2 w3 w_class
save mnistclassify_error test_err test_crerr train_err train_crerr;
end
% Version 1.000
%
% Code provided by Ruslan Salakhutdinov and Geoff Hinton
%
% Permission is granted for anyone to copy, use, modify, or distribute this
% program and accompanying programs and documents for any purpose, provided
% this copyright notice is retained and prominently displayed, along with
% a note saying that the original programs are available from our
% web page.
% The programs and documents are distributed without any warranty, express or
% implied. As the programs were written for research purposes only, they have
% not been tested to the degree that would be advisable in any important
% application. All use of these programs is entirely at the user's own risk.
% This program fine-tunes an autoencoder with backpropagation.
% Weights of the autoencoder are going to be saved in mnist_weights.mat
% and trainig and test reconstruction errors in mnist_error.mat
% You can also set maxepoch, default value is 200 as in our paper.
maxepoch=2; %maxepoch=200;
fprintf(1,'\nTraining discriminative model on MNIST by minimizing cross entropy error. \n');
fprintf(1,'60 batches of 1000 cases each. \n');
load mnistvhclassify
load mnisthpclassify
load mnisthp2classify
makebatches;
[numcases numdims numbatches]=size(batchdata); % 100 784 600
N=numcases; % 100
%%%% PREINITIALIZE WEIGHTS OF THE DISCRIMINATIVE MODEL%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
w1=[vishid; hidrecbiases]; % 784 250 + 1 250 = 785 250
w2=[hidpen; penrecbiases]; % 250 250 + 1 250 = 251 250
w3=[hidpen2; penrecbiases2]; % 250 500 + 1 500 = 251 500
w_class = 0.1*randn(size(w3,2)+1,10); % randn(501,10) = 501 10
%%%%%%%%%% END OF PREINITIALIZATION OF WEIGHTS %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
l1=size(w1,1)-1; % 784
l2=size(w2,1)-1; % 250
l3=size(w3,1)-1; % 250
l4=size(w_class,1)-1; % 500
l5=10;
test_err=[];
train_err=[];
for epoch = 1:maxepoch
%%%%%%%%%%%%%%%%%%%% COMPUTE TRAINING MISCLASSIFICATION ERROR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
err=0;
err_cr=0;
counter=0;
[numcases numdims numbatches]=size(batchdata); % 100 784 600
N=numcases; % 100
for batch = 1:numbatches % 1 : 600
data = [batchdata(:,:,batch)]; % 100 784
target = [batchtargets(:,:,batch)]; % 100 10 600
data = [data ones(N,1)]; % 100 785
w1probs = 1./(1 + exp(-data*w1)); w1probs = [w1probs ones(N,1)]; % 100 785 * 785 250 += 100 251
w2probs = 1./(1 + exp(-w1probs*w2)); w2probs = [w2probs ones(N,1)]; % 100 251 * 251 250 += 100 251
w3probs = 1./(1 + exp(-w2probs*w3)); w3probs = [w3probs ones(N,1)]; % 100 251 * 251 500 += 100 501
targetout = exp(w3probs*w_class); % 100 501 * 501 10 = 100 10
targetout = targetout./repmat(sum(targetout,2),1,10); % 100 10 = 100 10 ./ repmat ( 100 1), 1 10) = 100 10
[I J]=max(targetout,[],2); % 100 1 , 100 1 = 100 1 -->I has the value J has the sequence
[I1 J1]=max(target,[],2); % max(100 10,[],2) 100 1
counter=counter+length(find(J==J1)); % =6 for the first batch
err_cr = err_cr- sum(sum( target(:,1:end).*log(targetout))) ; %cross entrophy
end
train_err(epoch)=(numcases*numbatches-counter); % total number of errors for all the batches in this epoche
train_crerr(epoch)=err_cr/numbatches; % total cross enthropy error for the complete batchdata in this epoche
%%%%%%%%%%%%%% END OF COMPUTING TRAINING MISCLASSIFICATION ERROR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%% COMPUTE TEST MISCLASSIFICATION ERROR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
err=0;
err_cr=0;
counter=0;
[testnumcases testnumdims testnumbatches]=size(testbatchdata); % 100 784 100
N=testnumcases; % 100
for batch = 1:testnumbatches % 1: 100
data = [testbatchdata(:,:,batch)]; % 100 784
target = [testbatchtargets(:,:,batch)]; % 100 10 = (100 10 100(:,:,batch)
data = [data ones(N,1)]; % 100 785
w1probs = 1./(1 + exp(-data*w1)); w1probs = [w1probs ones(N,1)]; % 100 785 *785 250 = 100 250 -> 100 251
w2probs = 1./(1 + exp(-w1probs*w2)); w2probs = [w2probs ones(N,1)]; % 100 251 * 251 250 = 100 250 -> 100 251
w3probs = 1./(1 + exp(-w2probs*w3)); w3probs = [w3probs ones(N,1)]; % 100 251 * 251 50 = 100 50 -> 100 51
targetout = exp(w3probs*w_class); % 100 51 * 51 10 = 100 10
targetout = targetout./repmat(sum(targetout,2),1,10); % = 100 10 ./ repmat ( 100 1), 1 10) = 100 10
[I J]=max(targetout,[],2); % 100 1 , 100 1 = 100 1 -->I has the value J has the sequence
[I1 J1]=max(target,[],2); % max(100 10,[],2) 100 1
counter=counter+length(find(J==J1)); % =9 for the first batch
err_cr = err_cr- sum(sum( target(:,1:end).*log(targetout))); %cross entrophy
end
test_err(epoch)=(testnumcases*testnumbatches-counter); % total number of errors for all the batches in this epoche
test_crerr(epoch)=err_cr/testnumbatches; % total cross enthropy error for the complete batchdata in this epoche
fprintf(1,'Before epoch %d Train # misclassified: %d (from %d). Test # misclassified: %d (from %d) \t \t \n',...
epoch,train_err(epoch),numcases*numbatches,test_err(epoch),testnumcases*testnumbatches);
%%%%%%%%%%%%%% END OF COMPUTING TEST MISCLASSIFICATION ERROR %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
tt=0;
for batch = 1:numbatches/10
fprintf(1,'epoch %d batch %d\n',epoch,batch);
%%%%%%%%%%% COMBINE 10 MINIBATCHES INTO 1 LARGER MINIBATCH %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
tt=tt+1;
data=[];
targets=[];
for kk=1:10
data=[data
batchdata(:,:,(tt-1)*10+kk)]; % 1000 784
targets=[targets
batchtargets(:,:,(tt-1)*10+kk)]; % 1000 10
end
%%%%%%%%%%%%%%% PERFORM CONJUGATE GRADIENT WITH 3 LINESEARCHES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
max_iter=3;
if epoch<2 6="" fixed.="" holding="" irst="" nbsp="" original="" other="" p="" top-level="" update="" weights=""> N = size(data,1); % 1000 /1000 784)
XX = [data ones(N,1)]; % 1000 785
w1probs = 1./(1 + exp(-XX*w1)); w1probs = [w1probs ones(N,1)]; % 1000 785 * 785 250 = 1000 250 -> 1000 251
w2probs = 1./(1 + exp(-w1probs*w2)); w2probs = [w2probs ones(N,1)]; % 1000 251 * 251 250 = 1000 250 -> 1000 251
w3probs = 1./(1 + exp(-w2probs*w3)); %w3probs = [w3probs ones(N,1)]; % 1000 251 * 251 50 = 1000 50 -> 1000 51
VV = [w_class(:)']'; % 51 10 = 510 1
Dim = [l4; l5]; % 2 1
[X, fX] = minimize(VV,'CG_CLASSIFY_INIT',max_iter,Dim,w3probs,targets); % 510 1, 4 1 = min(510 1,'CG_CLASS...',3, 2 1, 1000 51, 1000 10
w_class = reshape(X,l4+1,l5); %reshape(X,51,10)= 51 10
else
VV = [w1(:)' w2(:)' w3(:)' w_class(:)']'; % 272060 1
Dim = [l1; l2; l3; l4; l5]; % 5 1
[X, fX] = minimize(VV,'CG_CLASSIFY',max_iter,Dim,data,targets); %[% 272060 1, 4 1]=mini..(272060 1,'CG_CLA..', 3, 5 1, 1000 784, 1000 10);
w1 = reshape(X(1:(l1+1)*l2),l1+1,l2); %reshape(272060(1 : 785*250,785,250) = 785 250
xxx = (l1+1)*l2; % 785 * 250
w2 = reshape(X(xxx+1:xxx+(l2+1)*l3),l2+1,l3); % reshape(X(251 : 251 * 250), 251, 250); = 251 250
xxx = xxx+(l2+1)*l3; % 785 * 250 + 251 * 250
w3 = reshape(X(xxx+1:xxx+(l3+1)*l4),l3+1,l4); % reshape(X(xxx+1: xxx+251*50, 251,50) = 251, 50
xxx = xxx+(l3+1)*l4; % 271550
w_class = reshape(X(xxx+1:xxx+(l4+1)*l5),l4+1,l5); % 51 10
end
%%%%%%%%%%%%%%% END OF CONJUGATE GRADIENT WITH 3 LINESEARCHES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
end
save mnistclassify_weights w1 w2 w3 w_class
save mnistclassify_error test_err test_crerr train_err train_crerr;
end
mnistclassify analysis 1
% Version 1.000
%
% Code provided by Ruslan Salakhutdinov and Geoff Hinton
%
% Permission is granted for anyone to copy, use, modify, or distribute this
% program and accompanying programs and documents for any purpose, provided
% this copyright notice is retained and prominently displayed, along with
% a note saying that the original programs are available from our
% web page.
% The programs and documents are distributed without any warranty, express or
% implied. As the programs were written for research purposes only, they have
% not been tested to the degree that would be advisable in any important
% application. All use of these programs is entirely at the user's own risk.
% This program pretrains a deep autoencoder for MNIST dataset
% You can set the maximum number of epochs for pretraining each layer
% and you can set the architecture of the multilayer net.
clear all
close all
maxepoch=1; % maxepoch=50;
numhid=250; numpen=250; numpen2=50; % numhid=500; numpen=500; numpen2=2000;
fprintf(1,'Converting Raw files into Matlab format \n');
%converter;
dos('erase *.ascii');
fprintf(1,'Pretraining a deep autoencoder. \n');
fprintf(1,'The Science paper used 50 epochs. This uses %3i \n', maxepoch);
makebatches;
[numcases numdims numbatches]=size(batchdata); % 100 784 600
fprintf(1,'Pretraining Layer 1 with RBM: %d-%d \n',numdims,numhid); % 784 250
restart=1;
rbm;
hidrecbiases=hidbiases; % 1 250
save mnistvhclassify vishid hidrecbiases visbiases; % 784 250, 1 250, 1 784
fprintf(1,'\nPretraining Layer 2 with RBM: %d-%d \n',numhid,numpen); % 250 250
batchdata=batchposhidprobs; % 100 250 600
numhid=numpen; %250
restart=1;
rbm;
hidpen=vishid; penrecbiases=hidbiases; hidgenbiases=visbiases;
save mnisthpclassify hidpen penrecbiases hidgenbiases; %hidpen 250 250, penrecbiases 1 250, hidgenbiases 1 250
fprintf(1,'\nPretraining Layer 3 with RBM: %d-%d \n',numpen,numpen2); % 250 50
batchdata=batchposhidprobs;
numhid=numpen2;
restart=1;
rbm;
hidpen2=vishid; penrecbiases2=hidbiases; hidgenbiases2=visbiases;
save mnisthp2classify hidpen2 penrecbiases2 hidgenbiases2; %hidpen2 250 50, penrecbiases2 1 50, hidgenbiases2 1 250
backpropclassify;
%
% Code provided by Ruslan Salakhutdinov and Geoff Hinton
%
% Permission is granted for anyone to copy, use, modify, or distribute this
% program and accompanying programs and documents for any purpose, provided
% this copyright notice is retained and prominently displayed, along with
% a note saying that the original programs are available from our
% web page.
% The programs and documents are distributed without any warranty, express or
% implied. As the programs were written for research purposes only, they have
% not been tested to the degree that would be advisable in any important
% application. All use of these programs is entirely at the user's own risk.
% This program pretrains a deep autoencoder for MNIST dataset
% You can set the maximum number of epochs for pretraining each layer
% and you can set the architecture of the multilayer net.
clear all
close all
maxepoch=1; % maxepoch=50;
numhid=250; numpen=250; numpen2=50; % numhid=500; numpen=500; numpen2=2000;
fprintf(1,'Converting Raw files into Matlab format \n');
%converter;
dos('erase *.ascii');
fprintf(1,'Pretraining a deep autoencoder. \n');
fprintf(1,'The Science paper used 50 epochs. This uses %3i \n', maxepoch);
makebatches;
[numcases numdims numbatches]=size(batchdata); % 100 784 600
fprintf(1,'Pretraining Layer 1 with RBM: %d-%d \n',numdims,numhid); % 784 250
restart=1;
rbm;
hidrecbiases=hidbiases; % 1 250
save mnistvhclassify vishid hidrecbiases visbiases; % 784 250, 1 250, 1 784
fprintf(1,'\nPretraining Layer 2 with RBM: %d-%d \n',numhid,numpen); % 250 250
batchdata=batchposhidprobs; % 100 250 600
numhid=numpen; %250
restart=1;
rbm;
hidpen=vishid; penrecbiases=hidbiases; hidgenbiases=visbiases;
save mnisthpclassify hidpen penrecbiases hidgenbiases; %hidpen 250 250, penrecbiases 1 250, hidgenbiases 1 250
fprintf(1,'\nPretraining Layer 3 with RBM: %d-%d \n',numpen,numpen2); % 250 50
batchdata=batchposhidprobs;
numhid=numpen2;
restart=1;
rbm;
hidpen2=vishid; penrecbiases2=hidbiases; hidgenbiases2=visbiases;
save mnisthp2classify hidpen2 penrecbiases2 hidgenbiases2; %hidpen2 250 50, penrecbiases2 1 50, hidgenbiases2 1 250
backpropclassify;
Subscribe to:
Posts (Atom)