cost_layers.py 文件源码

def _init_params(self):
        self.iBlocks = 1  # number of blocks in the input (from lower layer)

        W_em = self.init_fn(self.n_in,
                            self.n_class,
                            self.sparsity,
                            self.scale,
                            self.rng)
        self.W_em = theano.shared(W_em,
                                  name='W_%s' % self.name)
        self.b_em = theano.shared(
            self.bias_fn(self.n_class, self.bias_scale, self.rng),
            name='b_%s' % self.name)

        U_em = theano.shared(((self.rng.rand(self.iBlocks, self.n_class, 
            self.n_in, self.n_words_class)-0.5)/(self.n_words_class*self.n_in)
            ).astype(theano.config.floatX), name='U_%s'%self.name)
        self.U_em = U_em
        c_em = numpy.zeros((self.n_class, self.n_words_class), dtype='float32')
        n_words_last_class = self.n_out % self.n_words_class
        #c_em[-1, n_words_last_class:] = -numpy.inf
        self.c_em = theano.shared(c_em, name='c_%s' % self.name)

        self.params = [self.W_em, self.b_em, self.U_em, self.c_em]
        self.params_grad_scale = [self.grad_scale for x in self.params]