Hi， I met the same problem. Have you figured out how to solve it?
在 2015年12月23日星期三 UTC+8上午3:39:52，Hamid Reza Hassanzadeh写道： > > Hello, > Thanks, > No I'm doing regression. > > On Monday, December 21, 2015 at 3:12:05 PM UTC-5, nouiz wrote: >> >> Hi, >> >> you can use the HINT from the error message to find the exact line number >> where the problem come from. >> >> mse is useful when doing regression. so your target should be a vector, >> with only one value per example. Then the 2 vector will broadcast correctly >> together. >> >> Are you doing classification? If so, I should change your cost. If the >> class are ordered, then you can use them as a regression and use the mse >> cost, but you seem to use them as different not ordered class. >> >> Fred >> >> On Sun, Dec 20, 2015 at 8:41 PM, Hamid Reza Hassanzadeh < >> [email protected]> wrote: >> >>> Hi everyone, >>> I'm trying to find the weights of a neural network with mean square >>> error cost using gradient descent. Here is the theano function I create: >>> train_fn = theano.function( >>> inputs=[train_index], >>> outputs=self.finetune_cost, >>> updates=updates, >>> givens={ >>> self.x: dataset_x[train_index], >>> self.y: dataset_y[train_index] >>> } >>> ) >>> >>> with self.x = T.matrix('x') and self.y = T.dvector('y') and the >>> finetune_cost is defined as: >>> self.finetune_cost = self.regLayer.mean_sq_error(self.y) >>> >>> def mean_sq_error(self,y): >>> return T.mean((self.y_pred-y)**2) >>> >>> Now the problem is that when I compile I get the following error: >>> ValueError: Input dimension mis-match. (input[0].shape[1] = 1, >>> input[2].shape[1] = 46) >>> Apply node that caused the error: Elemwise{Composite{((i0 + i1) - >>> i2)}}[(0, 0)](Dot22.0, InplaceDimShuffle{x,0}.0, InplaceDimShuffle{x,0}.0) >>> Toposort index: 30 >>> Inputs types: [TensorType(float64, matrix), TensorType(float64, row), >>> TensorType(float64, row)] >>> Inputs shapes: [(46L, 1L), (1L, 1L), (1L, 46L)] >>> Inputs strides: [(8L, 8L), (8L, 8L), (368L, 8L)] >>> Inputs values: ['not shown', array([[ 0.]]), 'not shown'] >>> Outputs clients: [[Elemwise{sqr,no_inplace}(Elemwise{Composite{((i0 + >>> i1) - i2)}}[(0, 0)].0), Elemwise{Composite{((i0 * i1) / i2)}}[(0, >>> 1)](TensorConstant{(1L, 1L) of 2.0}, Elemwise{Composite{((i0 + i1) - >>> i2)}}[(0, 0)].0, Elemwise{mul,no_inplace}.0)]] >>> >>> HINT: Re-running with most Theano optimization disabled could give you a >>> back-trace of when this node was created. This can be done with by setting >>> the Theano flag 'optimizer=fast_compile'. If that does not work, Theano >>> optimizations can be disabled with 'optimizer=None'. >>> HINT: Use the Theano flag 'exception_verbosity=high' for a debugprint >>> and storage map footprint of this apply node. >>> >>> >>> I guess the problem is that self.y_pred is a column matrix whereas the y >>> is a dvector which is treated as a row matrix. What should I do? >>> >>> >>> >>> -- >>> >>> --- >>> You received this message because you are subscribed to the Google >>> Groups "theano-users" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to [email protected]. >>> For more options, visit https://groups.google.com/d/optout. >>> >> >> -- --- You received this message because you are subscribed to the Google Groups "theano-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
