This is a quick Mariana example to showcase:
In [2]:
#!kill -9 -1
!pip install git+https://github.com/Theano/Theano
!pip install git+https://github.com/Lasagne/Lasagne
!pip install git+https://github.com/tariqdaouda/Mariana
Collecting git+https://github.com/Theano/Theano
Cloning https://github.com/Theano/Theano to /tmp/pip-req-build-LImt5D
Requirement already satisfied: numpy>=1.9.1 in /usr/local/lib/python2.7/dist-packages (from Theano==1.0.2) (1.14.3)
Requirement already satisfied: scipy>=0.14 in /usr/local/lib/python2.7/dist-packages (from Theano==1.0.2) (0.19.1)
Requirement already satisfied: six>=1.9.0 in /usr/local/lib/python2.7/dist-packages (from Theano==1.0.2) (1.11.0)
Building wheels for collected packages: Theano
Running setup.py bdist_wheel for Theano ... - \ | / - \ | / - done
Stored in directory: /tmp/pip-ephem-wheel-cache-55Rdzf/wheels/64/f2/f4/6b1f50baf18aca2eab5d9b5a431b90e3d8be4711c8f7457eb7
Successfully built Theano
Installing collected packages: Theano
Successfully installed Theano-1.0.2
Collecting git+https://github.com/Lasagne/Lasagne
Cloning https://github.com/Lasagne/Lasagne to /tmp/pip-req-build-pocfxC
Requirement already satisfied: numpy in /usr/local/lib/python2.7/dist-packages (from Lasagne==0.2.dev1) (1.14.3)
Building wheels for collected packages: Lasagne
Running setup.py bdist_wheel for Lasagne ... - done
Stored in directory: /tmp/pip-ephem-wheel-cache-2482Ax/wheels/c4/20/90/9f7242c381402829c5918261e3eb51a87bc1d8521456749b57
Successfully built Lasagne
Installing collected packages: Lasagne
Successfully installed Lasagne-0.2.dev1
Collecting git+https://github.com/tariqdaouda/Mariana
Cloning https://github.com/tariqdaouda/Mariana to /tmp/pip-req-build-Bm1ZWE
Requirement already satisfied: theano in /usr/local/lib/python2.7/dist-packages (from Mariana==2.0.0rc1) (1.0.2)
Collecting pyGeno (from Mariana==2.0.0rc1)
Downloading https://files.pythonhosted.org/packages/2e/f8/3e7b1aa849e7100109559f0cc4b1ec872794cecf301d76db267dc67a6e1e/pyGeno-1.3.1.tar.gz (7.1MB)
100% |████████████████████████████████| 7.1MB 3.0MB/s
Collecting simplejson (from Mariana==2.0.0rc1)
Downloading https://files.pythonhosted.org/packages/8b/6c/c512c32124d1d2d67a32ff867bb3cdd5bfa6432660975f7ee753ed7ad886/simplejson-3.15.0.tar.gz (80kB)
100% |████████████████████████████████| 81kB 20.3MB/s
Requirement already satisfied: numpy in /usr/local/lib/python2.7/dist-packages (from Mariana==2.0.0rc1) (1.14.3)
Requirement already satisfied: Lasagne in /usr/local/lib/python2.7/dist-packages (from Mariana==2.0.0rc1) (0.2.dev1)
Requirement already satisfied: six>=1.9.0 in /usr/local/lib/python2.7/dist-packages (from theano->Mariana==2.0.0rc1) (1.11.0)
Requirement already satisfied: scipy>=0.14 in /usr/local/lib/python2.7/dist-packages (from theano->Mariana==2.0.0rc1) (0.19.1)
Collecting rabaDB>=1.0.4 (from pyGeno->Mariana==2.0.0rc1)
Downloading https://files.pythonhosted.org/packages/11/10/e8397c370efd954209089319b3d842a1689499a5b5c20812e5094d32b83f/rabaDB-1.0.5.tar.gz
Building wheels for collected packages: Mariana, pyGeno, simplejson, rabaDB
Running setup.py bdist_wheel for Mariana ... - done
Stored in directory: /tmp/pip-ephem-wheel-cache-nnk3G1/wheels/1f/ca/1b/22d3ed05d19c6b64cb8bd1e926058a1cf23ede03cc3a0d2eda
Running setup.py bdist_wheel for pyGeno ... - \ | done
Stored in directory: /content/.cache/pip/wheels/7c/74/4a/0c0e7cc9694e01e46419e890cfaff79671ff70457cd9cfe7af
Running setup.py bdist_wheel for simplejson ... - \ done
Stored in directory: /content/.cache/pip/wheels/2c/96/fb/b63af7400da79753dcd2a3f9bf5e7a3010e8d0233844445c2c
Running setup.py bdist_wheel for rabaDB ... - done
Stored in directory: /content/.cache/pip/wheels/ad/ab/a0/860907dc9ef89c041fbb4627dc11d9c05de0808939b6539b3e
Successfully built Mariana pyGeno simplejson rabaDB
Installing collected packages: rabaDB, pyGeno, simplejson, Mariana
Successfully installed Mariana-2.0.0rc1 pyGeno-1.3.1 rabaDB-1.0.5 simplejson-3.15.0
In [0]:
In [14]:
import Mariana.layers as ML
import Mariana.scenari as MS
import Mariana.costs as MC
import Mariana.activations as MA
import Mariana.regularizations as MR
import Mariana.settings as MSET
import numpy
ls = MS.GradientDescent(lr = 0.01, momentum=0.9)
cost = MC.NegativeLogLikelihood()
inp = ML.Input(28*28, name = "InputLayer")
h1 = ML.Hidden(300, activation = MA.ReLU(), name = "Hidden1", regularizations = [ MR.L1(0.0001) ])
h2 = ML.Hidden(300, activation = MA.ReLU(), name = "Hidden2", regularizations = [ MR.L1(0.0001) ])
o = ML.SoftmaxClassifier(10, learningScenari = [ls], cost = cost, name = "Probabilities1")
o.addNote("This is a note on the output layer", "I am the output layer")
#Connecting layers
inp > h1 > h2
concat = ML.C([inp, h2], name="Concatenation")
MLP_skip = concat > o
MLP_skip.addNote("This is a note on the model", "This is the skip connection example. Notes make it easier to collaborate")
MLP_skip.init()
>|\/| /-\ |-> | /-\ |\| /-\>
In [0]:
print "->train:\n", MLP_skip["Probabilities"].train({"InputLayer.inputs": [numpy.ones(784)], "Probabilities.targets": [1]})
print "---"
print "->test:\n", MLP_skip["Probabilities"].test({"InputLayer.inputs": [numpy.ones(784)], "Probabilities.targets": [1]})
print "---"
print "->the output of the first hidden layer, on the stream 'train' (has regularisations):\n", MLP_skip["Hidden1"].propagate["train"]({"InputLayer.inputs": [numpy.ones(784)]})
print "---"
print "->the output of the last layer, on the stream 'test' (no regularisations):\n", MLP_skip["Probabilities"].propagate["test"]({"InputLayer.inputs": [numpy.ones(784)]})
print "---"
#mix-in function: output on the second hidden + test score
f = MLP_skip["Hidden2"].propagate["test"] + MLP_skip["Probabilities"].test
print "->mix-in outputs:\n", f({"InputLayer.inputs": [numpy.ones(784)], "Probabilities.targets": [1]})
In [15]:
ls = MS.GradientDescent(lr = 0.01, momentum=0.9)
cost = MC.NegativeLogLikelihood()
inp = ML.Input(28*28, name = "InputLayer")
h1 = ML.Hidden(300, activation = MA.ReLU(), name = "Hidden1", regularizations = [ MR.L1(0.0001) ])
h2 = ML.Hidden(300, activation = MA.ReLU(), name = "Hidden2", regularizations = [ MR.L1(0.0001) ])
o1 = ML.SoftmaxClassifier(10, learningScenari = [ls], cost = cost, name = "Probabilities1")
o2 = ML.SoftmaxClassifier(3, learningScenari = [ls], cost = cost, name = "Probabilities2")
#Connecting layers
inp > h1 > h2
concat = ML.C([inp, h2], name="Concatenation")
h1 > o1
MLP_skip = concat > o2
MLP_skip.addNote("This is a note on the model", "This model has two outputs")
MLP_skip.init()
>|\/| /-\ |-> | /-\ |\| /-\>
In [28]:
print "->train on fisrt output:\n", MLP_skip["Probabilities1"].train({"InputLayer.inputs": [numpy.ones(784)], "Probabilities1.targets": [1]})
print "---"
#mix-in function: train on both outputs
f = MLP_skip["Probabilities1"].train + MLP_skip["Probabilities2"].train
print "->mix-in outputs:\n", f({"InputLayer.inputs": [numpy.ones(784)], "Probabilities1.targets": [1], "Probabilities2.targets": [0]})
->train on fisrt output:
OrderedDict([('Probabilities1.drive.train', array(9.08210724e-06))])
---
->mix-in outputs:
OrderedDict([('Probabilities1.drive.train', array(2.71914144e-06)), ('Probabilities2.drive.train', array(3.11565855))])
In [30]:
#Get the gradients for debugging and interpretation
MLP_skip["Probabilities1"].train.getGradients({"InputLayer.inputs": [numpy.ones(784)], "Probabilities1.targets": [1]})
Out[30]:
OrderedDict([('Hidden1.b',
array([ 0.00000000e+00, 0.00000000e+00, -1.51440640e-07, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -2.11944325e-07, 0.00000000e+00,
-8.45051287e-08, -3.29868775e-07, 0.00000000e+00, 0.00000000e+00,
-1.93888435e-08, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, -2.65512340e-07, 8.64366131e-08, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, -1.24948273e-08,
-2.46019679e-07, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, -1.22970078e-08, 0.00000000e+00, 1.97720300e-07,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -1.65620081e-07, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -1.32481057e-07, 0.00000000e+00,
0.00000000e+00, -6.08194523e-08, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 5.81935868e-08, -5.84462816e-08, -1.56930349e-07,
8.60929811e-09, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
4.93658236e-08, -1.84526200e-07, 0.00000000e+00, -3.14566429e-07,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 7.00792790e-08,
0.00000000e+00, 1.26540198e-07, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, -3.93229324e-08, 0.00000000e+00, -5.72147911e-07,
-2.27079875e-07, 0.00000000e+00, 4.10745621e-08, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -2.02455513e-07, -1.72548647e-07,
-1.75144999e-07, 0.00000000e+00, 0.00000000e+00, -1.89935493e-07,
0.00000000e+00, -1.86882820e-07, 0.00000000e+00, 0.00000000e+00,
-1.27546771e-08, -2.11789876e-07, 0.00000000e+00, -4.57477516e-07,
-3.43445307e-07, 2.19564985e-08, -5.97082067e-07, -3.61617158e-08,
0.00000000e+00, 0.00000000e+00, -2.37156055e-07, -1.34966324e-07,
0.00000000e+00, -7.44587445e-08, -1.26109273e-07, 0.00000000e+00,
-3.47191760e-07, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, -1.95023644e-07, -1.58811511e-08, 0.00000000e+00,
-6.75413584e-08, 0.00000000e+00, 0.00000000e+00, -4.94527696e-07,
-2.48081718e-07, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, -3.78809062e-07,
-2.53247795e-07, 3.32523530e-08, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 2.17325775e-07, 5.06991380e-09, 0.00000000e+00,
0.00000000e+00, 1.18529806e-07, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 7.00818330e-08,
-1.53581990e-07, -1.81760391e-07, 0.00000000e+00, 0.00000000e+00,
-4.86779420e-07, -2.39937422e-07, -1.76282489e-07, -1.44248390e-07,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 3.36340071e-08,
0.00000000e+00, 0.00000000e+00, 2.13194344e-07, -2.71212920e-08,
0.00000000e+00, -1.05014579e-07, 0.00000000e+00, -1.26743024e-07,
0.00000000e+00, 0.00000000e+00, -2.03868213e-07, -1.45612965e-07,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 8.54931305e-08,
0.00000000e+00, 0.00000000e+00, -3.29694996e-07, 0.00000000e+00,
-6.68591480e-08, 3.53386754e-08, 0.00000000e+00, -1.52685408e-07,
0.00000000e+00, 0.00000000e+00, -8.33013923e-08, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 1.75120233e-07, -6.74309758e-08,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, -2.09556698e-07, 0.00000000e+00, 0.00000000e+00,
-1.35095698e-07, 0.00000000e+00, -3.09324895e-07, 0.00000000e+00,
-1.67404823e-07, 0.00000000e+00, -2.82500020e-07, -4.09836435e-08,
3.20681591e-08, 0.00000000e+00, -3.76968672e-07, -2.83455635e-07,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
-3.99860741e-07, -5.43957936e-08, 0.00000000e+00, -8.75045694e-08,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -2.33373539e-07, 0.00000000e+00,
8.98120809e-08, 0.00000000e+00, -1.09677008e-07, 0.00000000e+00,
-1.79434179e-07, 0.00000000e+00, -2.32020621e-07, -1.04827456e-07,
0.00000000e+00, 0.00000000e+00, -2.58646132e-07, -2.14802375e-07,
-3.39904721e-07, 0.00000000e+00, -1.30528559e-07, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
-7.31538990e-09, 0.00000000e+00, 0.00000000e+00, -1.25133696e-07,
-2.43051234e-07, -4.61787118e-07, 0.00000000e+00, 0.00000000e+00,
7.23666588e-08, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
4.70579691e-08, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, -2.50880464e-07, -2.90441737e-08, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -5.18450782e-07, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, -4.44542484e-07, 1.57389788e-07, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 2.28660170e-07, 0.00000000e+00,
0.00000000e+00, -3.65161484e-07, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 2.71305695e-07, -1.06095190e-08, 0.00000000e+00,
-4.13253672e-07, 0.00000000e+00, 2.72072884e-07, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00])),
('Probabilities1.W',
array([[ 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, ...,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
[ 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, ...,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
[ 1.32662808e-07, -3.55292520e-06, 1.08867157e-06, ...,
5.67285648e-07, 1.79349404e-07, 5.92238387e-07],
...,
[ 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, ...,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
[ 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, ...,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
[ 0.00000000e+00, 0.00000000e+00, 0.00000000e+00, ...,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00]])),
('Hidden1.W',
array([[-0.0001 , -0.0001 , -0.00010015, ..., 0.0001 ,
-0.0001 , -0.0001 ],
[ 0.0001 , -0.0001 , -0.00010015, ..., -0.0001 ,
0.0001 , -0.0001 ],
[-0.0001 , -0.0001 , -0.00010015, ..., -0.0001 ,
-0.0001 , 0.0001 ],
...,
[-0.0001 , 0.0001 , -0.00010015, ..., -0.0001 ,
0.0001 , -0.0001 ],
[ 0.0001 , -0.0001 , -0.00010015, ..., -0.0001 ,
-0.0001 , 0.0001 ],
[-0.0001 , -0.0001 , -0.00010015, ..., -0.0001 ,
0.0001 , -0.0001 ]])),
('Probabilities1.b',
array([ 1.01529987e-07, -2.71913775e-06, 8.33186122e-07, 7.39746017e-08,
8.77679988e-08, 5.05009588e-07, 9.29979193e-08, 4.34157133e-07,
1.37260344e-07, 4.53254054e-07]))])
In [29]:
#Get the updates for debugging and interpretation
MLP_skip["Probabilities1"].train.getUpdates({"InputLayer.inputs": [numpy.ones(784)], "Probabilities1.targets": [1]})
Out[29]:
OrderedDict([('Hidden1.b',
array([ 0.00000000e+00, 0.00000000e+00, 1.34647880e-03, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 8.48849536e-04, 0.00000000e+00,
-4.02560261e-05, 2.22665159e-03, 0.00000000e+00, 0.00000000e+00,
-1.96807503e-04, 0.00000000e+00, -1.69743639e-03, 0.00000000e+00,
0.00000000e+00, 2.17963700e-03, -1.07524098e-03, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -2.59541859e-03, -1.61847860e-03,
2.26993545e-03, -1.04499160e-03, -1.73618232e-03, 0.00000000e+00,
0.00000000e+00, -2.18082692e-04, 0.00000000e+00, -1.88735617e-03,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 9.94986334e-04, 0.00000000e+00,
0.00000000e+00, -9.18299213e-04, 1.02780545e-04, 0.00000000e+00,
0.00000000e+00, -1.23485384e-04, 0.00000000e+00, -2.05070289e-03,
0.00000000e+00, -9.56866453e-04, -3.35028715e-04, 1.61693060e-03,
-1.20825866e-04, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
6.64830328e-04, 7.97307791e-04, 0.00000000e+00, 2.18433587e-03,
0.00000000e+00, 0.00000000e+00, -1.86547522e-03, -1.36430322e-03,
0.00000000e+00, -1.27746298e-03, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, -9.65450793e-04, 0.00000000e+00, 2.84429050e-03,
7.79130317e-04, -2.09953467e-03, -4.33320229e-04, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -2.25217849e-04, 6.49607924e-04,
8.30783556e-04, 0.00000000e+00, 0.00000000e+00, 1.69091171e-03,
0.00000000e+00, 1.16761124e-03, -2.91348748e-03, 0.00000000e+00,
4.03448998e-04, 2.42187042e-04, 0.00000000e+00, 3.31536738e-03,
2.62294508e-03, -1.19767344e-04, 3.12220519e-03, 6.08884319e-04,
-2.60828442e-03, -3.10020593e-03, 2.60553765e-03, -2.15235143e-05,
0.00000000e+00, 4.51284505e-04, -1.02537862e-04, 0.00000000e+00,
1.94051025e-03, -1.90192150e-03, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 9.21169254e-04, -4.96632337e-04, 0.00000000e+00,
1.01029547e-03, 0.00000000e+00, 0.00000000e+00, 1.41269408e-03,
1.10460553e-03, 0.00000000e+00, -3.17678100e-03, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 2.90763185e-03,
1.13008946e-03, 9.26344540e-04, -1.04658351e-03, -2.08606178e-03,
0.00000000e+00, -7.19384722e-04, 1.07522502e-03, -2.74519214e-03,
0.00000000e+00, -1.02028108e-03, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, -1.91227714e-03,
-1.78764971e-04, 1.41518995e-03, 0.00000000e+00, 0.00000000e+00,
2.96766872e-03, 1.48083560e-03, 1.19553234e-03, 7.18897373e-04,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, -5.27228553e-04,
0.00000000e+00, 0.00000000e+00, -2.12193559e-03, 2.94555924e-04,
0.00000000e+00, 1.20982687e-03, 0.00000000e+00, 1.19083139e-03,
0.00000000e+00, 0.00000000e+00, 5.55178807e-04, 1.41555877e-03,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -2.01147885e-03, 5.81348298e-04,
0.00000000e+00, 0.00000000e+00, 2.98715488e-03, 0.00000000e+00,
1.42425901e-03, -8.27788152e-05, 0.00000000e+00, 7.72222135e-04,
0.00000000e+00, 0.00000000e+00, 8.01247478e-04, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -7.66115096e-04, -7.82680950e-04,
-1.51568017e-03, 0.00000000e+00, -1.22228509e-03, 0.00000000e+00,
0.00000000e+00, 5.50177632e-04, -3.17134989e-03, 0.00000000e+00,
-1.75817537e-04, 0.00000000e+00, 1.13000141e-03, 0.00000000e+00,
8.61101732e-04, -1.21781216e-03, 1.65067116e-03, 2.66864417e-04,
-1.09295624e-03, 0.00000000e+00, 1.79242970e-03, 1.65655799e-03,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
2.45448832e-03, 1.86823925e-04, 0.00000000e+00, -5.72054779e-05,
-1.15094897e-03, 0.00000000e+00, -3.34405186e-03, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -9.75154770e-04, 0.00000000e+00,
-5.56922067e-04, -4.40392389e-04, 1.29051488e-03, -1.48277189e-03,
1.94121839e-03, 0.00000000e+00, 1.74576733e-03, 1.02305724e-03,
0.00000000e+00, 0.00000000e+00, 1.93963213e-03, 1.02541851e-03,
1.60393944e-03, -1.28470931e-03, 9.60136756e-04, 0.00000000e+00,
-1.68770815e-03, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
-9.76542690e-04, 0.00000000e+00, 0.00000000e+00, -1.01681949e-04,
8.52005559e-04, 2.73401191e-03, -8.72021963e-04, 0.00000000e+00,
-5.64307753e-05, 0.00000000e+00, 0.00000000e+00, -3.05852029e-03,
-1.49437639e-03, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
-2.80414420e-03, 4.31460784e-04, -1.58136825e-03, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 2.50055798e-03, 0.00000000e+00,
0.00000000e+00, -1.40282969e-03, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 1.11020948e-03, -8.60257371e-04, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -2.35308981e-03, 0.00000000e+00,
0.00000000e+00, 5.66017806e-04, 0.00000000e+00, 0.00000000e+00,
-2.15030922e-03, -1.09800252e-03, -1.27551021e-04, -2.35079730e-03,
2.12463431e-03, 0.00000000e+00, -1.51184543e-03, -1.65298268e-03,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00])),
('Hidden1.b.momentum',
array([ 0.00000000e+00, 0.00000000e+00, -9.07307091e-03, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -5.77540960e-03, 0.00000000e+00,
1.91361055e-04, -1.50533589e-02, 0.00000000e+00, 0.00000000e+00,
1.30471102e-03, 0.00000000e+00, 1.07004881e-02, 0.00000000e+00,
0.00000000e+00, -1.46039444e-02, 7.21588214e-03, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 1.74581530e-02, 1.06435899e-02,
-1.52925051e-02, 6.86585400e-03, 1.16183629e-02, 0.00000000e+00,
0.00000000e+00, 1.45171829e-03, 0.00000000e+00, 1.27046072e-02,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -6.73745573e-03, 0.00000000e+00,
0.00000000e+00, 5.34819602e-03, -7.59250975e-04, 0.00000000e+00,
0.00000000e+00, 7.46641533e-04, 0.00000000e+00, 1.36468358e-02,
0.00000000e+00, 6.40614540e-03, 2.15335988e-03, -1.07415446e-02,
8.23274151e-04, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
-4.32382711e-03, -5.47382903e-03, 0.00000000e+00, -1.48246118e-02,
0.00000000e+00, 0.00000000e+00, 1.25864349e-02, 9.12409204e-03,
0.00000000e+00, 8.69962730e-03, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 6.31276617e-03, 0.00000000e+00, -1.94157858e-02,
-5.32896658e-03, 1.40829452e-02, 2.96906506e-03, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 1.38698910e-03, -4.43303494e-03,
-5.67770681e-03, 0.00000000e+00, 0.00000000e+00, -1.13589356e-02,
0.00000000e+00, -7.97088523e-03, 1.96650648e-02, 0.00000000e+00,
-2.72307949e-03, -1.76051099e-03, 0.00000000e+00, -2.23982229e-02,
-1.76001444e-02, 8.24574515e-04, -2.12137332e-02, -4.08889010e-03,
1.74541586e-02, 2.08648545e-02, -1.74488404e-02, 1.73481350e-06,
0.00000000e+00, -3.02483366e-03, 6.27642236e-04, 0.00000000e+00,
-1.31027259e-02, 1.27289393e-02, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, -6.25095625e-03, 3.34304062e-03, 0.00000000e+00,
-6.76403440e-03, 0.00000000e+00, 0.00000000e+00, -9.87366618e-03,
-7.59051850e-03, 0.00000000e+00, 2.12898804e-02, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, -1.95440493e-02,
-7.69093410e-03, -6.03931171e-03, 6.62016198e-03, 1.40313942e-02,
0.00000000e+00, 5.00667052e-03, -7.14379427e-03, 1.82427877e-02,
0.00000000e+00, 7.00382489e-03, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.27101740e-02,
1.04664614e-03, -9.52725613e-03, 0.00000000e+00, 0.00000000e+00,
-2.00719867e-02, -9.96009439e-03, -8.08878112e-03, -4.88963686e-03,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 3.52354994e-03,
0.00000000e+00, 0.00000000e+00, 1.43425315e-02, -1.91304985e-03,
0.00000000e+00, -8.16783457e-03, 0.00000000e+00, -7.90972297e-03,
0.00000000e+00, 0.00000000e+00, -3.91039687e-03, -9.50454431e-03,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 1.35019165e-02, -3.79187488e-03,
0.00000000e+00, 0.00000000e+00, -2.00539933e-02, 0.00000000e+00,
-9.47561570e-03, 6.36504792e-04, 0.00000000e+00, -5.26895062e-03,
0.00000000e+00, 0.00000000e+00, -5.40412735e-03, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 5.24618011e-03, 5.13972498e-03,
1.02492474e-02, 0.00000000e+00, 7.70928376e-03, 0.00000000e+00,
0.00000000e+00, -3.86676002e-03, 2.10325178e-02, 0.00000000e+00,
1.09914753e-03, 0.00000000e+00, -7.70850877e-03, 0.00000000e+00,
-5.87465732e-03, 7.96870452e-03, -1.11797218e-02, -1.80003682e-03,
7.26820131e-03, 0.00000000e+00, -1.21888192e-02, -1.12437091e-02,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
-1.66393617e-02, -1.26427918e-03, 0.00000000e+00, 3.41353809e-04,
7.65315423e-03, 0.00000000e+00, 2.23011389e-02, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 6.28689397e-03, 0.00000000e+00,
3.80296555e-03, 2.93258248e-03, -8.63148436e-03, 1.00621692e-02,
-1.30036846e-02, 0.00000000e+00, -1.17436526e-02, -6.86236701e-03,
0.00000000e+00, 0.00000000e+00, -1.31072146e-02, -7.02233910e-03,
-1.09039601e-02, 8.74511404e-03, -6.51182845e-03, 0.00000000e+00,
1.13274234e-02, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
6.40680790e-03, 0.00000000e+00, 0.00000000e+00, 4.80586297e-04,
-5.83379995e-03, -1.85624197e-02, 5.86391182e-03, 0.00000000e+00,
4.68465128e-04, 0.00000000e+00, 0.00000000e+00, 2.05088168e-02,
9.90026344e-03, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
1.89780027e-02, -3.08836288e-03, 1.04654648e-02, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, -1.70380367e-02, 0.00000000e+00,
0.00000000e+00, 9.39745661e-03, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, -7.74312744e-03, 5.85546781e-03, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 1.58612556e-02, 0.00000000e+00,
0.00000000e+00, -3.99866501e-03, 0.00000000e+00, 0.00000000e+00,
1.46049803e-02, 7.53740910e-03, 8.84645762e-04, 1.58088702e-02,
-1.43920470e-02, 0.00000000e+00, 1.02976509e-02, 1.11472720e-02,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00])),
('Probabilities1.W',
array([[ 0.00924904, 0.04357116, 0.06890452, ..., 0.05960178,
0.0068972 , -0.11916905],
[ 0.23126429, -0.01205161, -0.06317142, ..., 0.03436202,
-0.0837746 , 0.0399785 ],
[-0.03191531, 0.06218873, 0.00520463, ..., -0.0138726 ,
-0.0694522 , -0.02506317],
...,
[ 0.11550306, 0.07997372, 0.09754261, ..., -0.0632434 ,
0.0910609 , 0.02819328],
[-0.02586889, 0.0998327 , 0.03985503, ..., 0.08392605,
0.1319304 , -0.03700615],
[-0.09042405, -0.03378866, 0.10019364, ..., -0.04561189,
-0.01316771, -0.05808974]])),
('Probabilities1.W.momentum',
array([[ 0. , 0. , 0. , ..., 0. ,
0. , 0. ],
[ 0. , 0. , 0. , ..., 0. ,
0. , 0. ],
[ 0.00571342, -0.04879519, 0.00884413, ..., 0.00338936,
0.00615124, 0.00151468],
...,
[ 0. , 0. , 0. , ..., 0. ,
0. , 0. ],
[ 0. , 0. , 0. , ..., 0. ,
0. , 0. ],
[ 0. , 0. , 0. , ..., 0. ,
0. , 0. ]])),
('Hidden1.W',
array([[-0.09203534, -0.0306802 , -0.00599801, ..., 0.03658268,
-0.05734369, -0.01589865],
[ 0.04604353, -0.04142277, -0.00041159, ..., -0.04508508,
0.0389231 , -0.00032717],
[-0.01911034, -0.06623062, -0.01788632, ..., -0.06593067,
-0.02994138, 0.00297791],
...,
[-0.00654239, 0.06390151, -0.04427453, ..., -0.01994633,
0.01366618, -0.00764931],
[ 0.0333654 , -0.00048473, -0.09249187, ..., -0.02849318,
-0.03834503, 0.00128915],
[-0.00230408, -0.03888115, -0.04023373, ..., -0.06812444,
0.11394296, -0.02966588]])),
('Hidden1.W.momentum',
array([[-6.86189404e-05, -6.86189404e-05, -9.14168985e-03, ...,
6.86189404e-05, -6.86189404e-05, -6.86189404e-05],
[ 6.86189404e-05, -6.86189404e-05, -9.14168985e-03, ...,
-6.86189404e-05, 6.86189404e-05, -6.86189404e-05],
[-6.86189404e-05, -6.86189404e-05, -9.14168985e-03, ...,
-6.86189404e-05, -6.86189404e-05, 6.86189404e-05],
...,
[-6.86189404e-05, 6.86189404e-05, -9.14168985e-03, ...,
-6.86189404e-05, 6.86189404e-05, -6.86189404e-05],
[ 6.86189404e-05, -6.86189404e-05, -9.14168985e-03, ...,
-6.86189404e-05, -6.86189404e-05, 6.86189404e-05],
[-6.86189404e-05, -6.86189404e-05, -9.14168985e-03, ...,
-6.86189404e-05, 6.86189404e-05, -6.86189404e-05]])),
('Probabilities1.b',
array([-0.0022705 , 0.01915674, -0.00327358, -0.00519217, -0.00133102,
-0.000805 , -0.00198406, -0.00127828, -0.00249152, -0.0005306 ])),
('Probabilities1.b.momentum',
array([ 0.01526163, -0.12944637, 0.0227595 , 0.03423116, 0.00902683,
0.00568669, 0.01329741, 0.0088057 , 0.01659171, 0.00378574]))])
Content source: tariqdaouda/Mariana
Similar notebooks: