Sage不可变向量错误

Question

我正在尝试在Sage中实现一些功能，但是我一直收到以下错误：

*Error in lines 38-53
Traceback (most recent call last):
  File "/projects/42e45a19-7a43-4495-8dcd-353625dfce66/.sagemathcloud/sage_server.py", line 879, in execute
    exec compile(block+'\n', '', 'single') in namespace, locals
  File "", line 13, in <module>
  File "sage/modules/vector_integer_dense.pyx", line 185, in sage.modules.vector_integer_dense.Vector_integer_dense.__setitem__ (build/cythonized/sage/modules/vector_integer_dense.c:3700)
    raise ValueError("vector is immutable; please change a copy instead (use copy())")
ValueError: vector is immutable; please change a copy instead (use copy())*

我已经准确定位了确切的位置（在while循环的末尾，"print 'marker 1'"和"print' marker 2 '之间的那一行，请参见下面的代码），看起来我不能从循环内部更改矩阵"weights"的条目（我在循环之前定义）。错误消息提示使用copy()函数，但我不知道这如何解决我的问题，因为我只会创建一个本地副本，而下一次循环不会得到这些更改后的值，对吗？所以，是否有人知道如何定义这个矩阵，以便我可以从循环内部进行更改？如果不可能，有人能解释一下吗？

感谢您的帮助。

---

代码：

m = 3  # Dimension of inputs to nodes
n = 1  # Dimension of output
v = 4  # Number of training vectors
r = 0.1 # Learning Rate
T = 10     # Number of iterations

# Input static Biases, i.e. sum must be smaller than this vector. For dynamic biases, set this vector to 0, increase m by one and set xi[0]=-1 for all inputs i (and start the acual input at xi[1])
bias = list(var('s_%d' % i) for i in range(n))
bias[0] = 0.5

# Input the training vectors and targets

x0 = list(var('s_%d' % i) for i in range(m))
x0[0]=1
x0[1]=0
x0[2]=0
target00=1

x1 = list(var('s_%d' % i) for i in range(m))
x1[0]=1
x1[1]=0
x1[2]=1
target10=1

x2 = list(var('s_%d' % i) for i in range(m))
x2[0]=1
x2[1]=1
x2[2]=0
target20=1

x3 = list(var('s_%d' % i) for i in range(m))
x3[0]=1
x3[1]=1
x3[2]=1
target30=0

targets = matrix(v,n,[[target00],[target10],[target20],[target30]])

g=matrix([x0,x1,x2,x3])
inputs=copy(g)

# Initialize weights, or leave at 0 (i.e.,change nothing)

weights=matrix(m,n)


print weights.transpose()

z = 0
a = list(var('s_%d' % j) for j in range(n))

while(z<T):
    Q = inputs*weights
    S = copy(Q)
    for i in range(v):
        y = copy(a)
        for j in range(n):
            if S[i][j] > bias[j]:
                y[j] = 1
            else:
                y[j] = 0
            for k in range(m):
                print 'marker 1'
                weights[k][j] = weights[k][j] + r*(targets[i][j]-y[j])*inputs[i][k]
                print 'marker 2'
    print weights.transpose

    z +=1

- Angela

Answer 1

这是Sage向量的一个基本属性 - 默认情况下它们是不可变的Python对象。

sage: M = matrix([[2,3],[3,2]])
sage: M[0][1] = 5
---------------------------------------------------------------------------
<snip>
ValueError: vector is immutable; please change a copy instead (use copy())

请注意，错误是由于向量是不可变的。这是因为你取了0行，它是一个向量（不可变、可哈希等）。

但如果你使用以下语法，应该就没问题了。

sage: M[0,1] = 5
sage: M
[2 5]
[3 2]

这里您直接修改了元素。希望这有所帮助，享受Sage吧！

- kcrisman