956ac38fd58a1ef65c18dffc06a22d2e628e3a16,GPflow/conditionals.py,,gaussian_gp_predict_whitened,#Any#Any#Any#Any#Any#,141

Before Change

    
 
    //compute kernel stuff
    num_data = X.shape[0]
    Kdiag = kern.Kdiag(Xnew)
    Kmn = kern.K(X, Xnew)
    Kmm = kern.K(X) + eye(num_data)*1e-4
    Lm = tf.cholesky(Kmm)

    //this is O(N M^2)
    A = tf.user_ops.triangular_solve(Lm, Kmn, "lower")

    //construct the mean and variance of q(f)
    fmean = tf.matmul(tf.transpose(A), q_mu)
    if q_sqrt.get_shape().ndims==2:
        //we hae a diagonal form for q(v)
        q_var = np.square(q_sqrt)
        //fvar = Kdiag[:,None] + tf.reduce_sum((tf.square(tf.transpose(A)))[:,:,None] * (q_var[None, :,:] - 1),1)
        fvar = tf.reshape(Kdiag, (-1,1)) + tf.reduce_sum(tf.reshape(tf.square(tf.transpose(A)), )[:,:,None] * (q_var[None, :,:] - 1),1)
    elif q_sqrt.get_shape().ndims ==3:
        // we have the cholesky form for q(v)
        fvar = Kdiag[:,None] - tf.reduce_sum(np.square(A), 0)[:,None]
        def f(w):
            R = tf.triu(w)
            RA = tf.matmul(R, A)
            return tf.square(RA).sum(0)

After Change

        fvar = tf.reshape(Kdiag, (-1,1)) + tf.reduce_sum(tf.reshape(tf.square(tf.transpose(A)), )[:,:,None] * (q_var[None, :,:] - 1),1)
    elif q_sqrt.get_shape().ndims ==3:
        // we have the cholesky form for q(v)
        fvar = Kdiag - tf.reduce_sum(np.square(A), 0)
        fvar = tf.expand_dims(fvar, 1)
        def f(w):
            R = tf.triu(w)
            RA = tf.matmul(R, A)
            return tf.square(RA).sum(0)
        projected_var, _ = theano.scan(f, q_sqrt.swapaxes(0,2))
        fvar += tf.transpose(projected_var)

    return fmean, fvar

Instances