在看GRAM源码，记录一下Theano的语法。参考：Theano教程系列，Theano API documentation, Theano教程

  1. 加载theano和numpy模块，并创建function:

    1
    2
    3
    import numpy as np
    import theano.tensor as T
    from theano import function

  2. 定义常量(scalar)及函数(function)：

    1
    2
    3
    4
    5
    6
    7
    8
    x = T.dscalar('x')
    y = T.dscalar('y')
    z = x+y

    f = function([x,y],z)

    print(f(2,3))
    # 5.0

  3. 打印原函数

    1
    2
    3
    4
    from theano import pp

    print(pp(z))
    # (x+y)

  4. 定义矩阵

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    x = T.dmatrix('x')
    y = T.dmatrix('y')
    z = x+y

    printf(
    	f(np.arange(12).reshape(3,4),10*np.ones(3,4))
    )

    '''
    [[ 10.  11.  12.  13.]
     [ 14.  15.  16.  17.]
     [ 18.  19.  20.  21.]]
     '''

  5. Function的用法

    1
    2
    3
    import numpy as np
    import theano.tensor as T
    import theano

  • 激活函数

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    x = T.dmatrix('x')
    s = 1/(1+T.exp(-x))

    logistic = theano.function([x],s)

    print(logistic([[0,1],[-2,-3]]))

    '''
    [[ 0.5         0.73105858]
     [ 0.26894142  0.11920292]]

  • 多输入/输出

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    a,b = T.dmatrices('a','b')

    diff = a-b
    abs_diff = abs(a-b)

    f = theano.function([a,b],[diff,abs_diff])

    x1,x2= f(
        np.ones((2,2)), # a
        np.arange(4).reshape((2,2))  # b
    )

    """
    array([[ 1.,  0.],
           [-1., -2.]]),
    array([[ 1.,  0.],
           [ 1.,  2.]]),  
    """

  • 默认值 & 名字

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    21
    22
    23
    24
    25
    x,y,w = T.dscalars('x','y','w')
    z = (x+y)*w

    f = theano.function([x,
                         theano.In(y,value=1),
                         theano.In(w,value=2)],
                        z)

    print(f(23))    # 使用默认
    print(f(23,1,4)) # 不使用默认
    """
    48.0
    100.0
    """

    f = theano.function([x,
                         theano.In(y,value=1),
                         theano.In(w,value=2,name='weights')],
                        z)
                        
    print (f(23,1,weights=4)) ##调用方式

    """
    100.0
    """

  • Shared变量

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    state = theano.shared(np.array(0,dtype=np.float64), 'state')     # inital state = 0
    inc = T.scalar('inc', dtype=state.dtype)
    accumulator = theano.function([inc], state, updates=[(state, state+inc)]) # updates: state+=inc

    # to get variable value
    print(state.get_value())
    # 0.0

    accumulator(1)   # return previous value, 0 in here
    print(state.get_value())
    # 1.0

    accumulator(10)  # return previous value, 1 in here
    print(state.get_value())
    # 11.0

    state.set_value(-1)
    accumulator(3)
    print(state.get_value())
    # 2.0

  • 激活函数：
    theano.tensor.nnet.nnet.sigmoid(x)
    softplus(), relu(), softmax(),tanh()

  • Layer类

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    l1 = Layer(inputs, in_size, out_size, T.nnet.relu)
    l2 = Layer(l1.outputs, in_size, out_size, None) # None采用默认的线性激活函数

    class Layer(object):
        def __init__(self, inputs, in_size, out_size, activation_function=None):
            self.W = theano.shared(np.random.normal(0, 1, (in_size, out_size)))
            self.b = theano.shared(np.zeros((out_size, )) + 0.1)
            self.Wx_plus_b = T.dot(inputs, self.W) + self.b
            self.activation_function = activation_function
            if activation_function is None:
                self.outputs = self.Wx_plus_b
            else:
                self.outputs = self.activation_function(self.Wx_plus_b)

  • 数据类型
    x=T.scalar(‘myvar’,dtype=theano.config.floatX)#创建0维阵列
    x=T.vector(‘myvar’,dtype=theano.config.floatX)#创建以为阵列
    x=T.row(‘myvar’,dtype=theano.config.floatX)#创建二维阵列,行数为1
    x=T.col(‘myvar’,dtype=theano.config.floatX)#创建二维阵列,列数为1
    x=T.matrix(‘myvar’,dtype=theano.config.floatX)#创建二维矩阵
    x=T.tensor3(‘myvar’,dtype=theano.config.floatX)#创建三维张量
    x=T.tensor4(‘myvar’,dtype=theano.config.floatX)#创建四维张量
    x=T.tensor5(‘myvar’,dtype=theano.config.floatX)#创建五维张量
    x.ndim#输出维度看看