chainerで2つの違うモーダルを共通空間に射影するときのモデルコード

model.py

# Network definition
class MLP(chainer.Chain):

    def __init__(self, n_units, n_out):
        super(MLP, self).__init__(
            # the size of the inputs to each layer will be inferred
            l1=L.Linear(None, n_units),  # n_in -> n_units
            l2=L.Linear(None, n_units),  # n_units -> n_units
            l3=L.Linear(None, n_out),  # n_units -> n_out
        )

    def __call__(self, x):
        h1 = F.relu(self.l1(x))
        h2 = F.relu(self.l2(h1))
        return self.l3(h2)

class CommonNet(chainer.Chain):
    def __init__(self, _model1, _model2):
        super(CommonNet, self).__init__(
            model1 = _model1,
            model2 = _model2,
        )

    def __call__(self, x, y):
        h1 = self.model1(x)
        h2 = self.model2(y)
        self.loss = F.mean_squared_error(h1, h2)
        reporter.report({'loss': self.loss}, self)
        return self.loss

enc1 = MLP(10,10)
enc2 = MLP(10,10)

model = CommonNet(enc1, enc2)

上記のようなモデルを組み，あとは，tupled_datasetで2つのデータを用意して，trainerでtrainさせればよい．