Multilayer perceptron with target variable as array instead of a single value

Question

I am new to deep learning and I have been trying to use the theano library to train my data. MLP tutorial here has a scalar output value while my use case has an array with a 1 corresponding to the value depicted in the output.

For example (assume the possible scalar values are 0,1,2,3,4,5),

0 = [1,0,0,0,0,0]
1 = [0,1,0,0,0,0]
2 = [0,0,1,0,0,0]

I have only modified the code to read my input and output (output now is a 2 dimensional array or matrix in the parlance of theano). Other parts of the code is as is from the MLP tutorial pasted above.

The error I am getting is in the following function

test_model = theano.function(inputs=[index],
            outputs=classifier.errors(y),
            givens={
                x: test_set_x[index * batch_size:(index + 1) * batch_size],
                y: test_set_y[index * batch_size:(index + 1) * batch_size]}) //line 286

Error stack:

  Traceback (most recent call last):
  File "mlp.py", line 398, in <module>
    test_mlp()
  File "mlp.py", line 286, in test_mlp
    y: test_set_y[index * batch_size:(index + 1) * batch_size]})
  File "/usr/local/lib/python2.7/dist-packages/theano/compile/function.py", line 223, in function
    profile=profile)
  File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py", line 490, in pfunc
    no_default_updates=no_default_updates)
  File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py", line 241, in rebuild_collect_shared
    cloned_v = clone_v_get_shared_updates(outputs, copy_inputs_over)
  File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py", line 92, in clone_v_get_shared_updates
    clone_a(v.owner, copy_inputs_over)
  File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py", line 131, in clone_a
    clone_v_get_shared_updates(i, copy_inputs_over)
  File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py", line 92, in clone_v_get_shared_updates
    clone_a(v.owner, copy_inputs_over)
  File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py", line 131, in clone_a
    clone_v_get_shared_updates(i, copy_inputs_over)
  File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py", line 92, in clone_v_get_shared_updates
    clone_a(v.owner, copy_inputs_over)
  File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py", line 135, in clone_a
    strict=rebuild_strict)
  File "/usr/local/lib/python2.7/dist-packages/theano/gof/graph.py", line 213, in clone_with_new_inputs
    new_inputs[i] = curr.type.filter_variable(new)
  File "/usr/local/lib/python2.7/dist-packages/theano/tensor/type.py", line 205, in filter_variable
    self=self)<br><br>
  TypeError: Cannot convert Type TensorType(int64, matrix) (of Variable Subtensor{int64:int64:}.0) into Type TensorType(int32, vector). You can try to manually convert Subtensor{int64:int64:}.0 into a TensorType(int32, vector).

I would like to know how to change this theano.function to accommodate the y value as a matrix.

Answer 1

You need to define y as T.imatrix() instead of T.lvector().