syna.functions.loss module

class syna.functions.loss.MeanSquaredError

Bases: Function

Mean squared error over the batch.

backward(gy)

forward(x0, x1)

class syna.functions.loss.SoftmaxCrossEntropy

Bases: Function

Softmax cross entropy (expects integer class labels t).

backward(gy)

forward(x, t)

syna.functions.loss.accuracy(y, t) → Tensor

Compute classification accuracy: mean(pred == t).

syna.functions.loss.mean_squared_error(x0, x1) → Tensor

Mean squared error between x0 and x1.

syna.functions.loss.mean_squared_error_simple(x0, x1)

Simple mean squared error (for convenience).

syna.functions.loss.softmax_cross_entropy(x, t) → Tensor

Softmax cross entropy between logits x and labels t.

syna.functions.loss.softmax_cross_entropy_simple(x, t) → Tensor

Simple softmax cross entropy using explicit softmax and log clipping.