cosine_distances

def cosine_distances(X, Y):
    """
    X : Target example score vector DataFrame with inst_id as the first column
    Y : All example score vector DataFrame with inst_id as the first column
    return pair-wise cosine distance DataFrame
    """
    from sklearn.metrics import pairwise
    x_header = X.iloc[:, 0].values
    X = X.iloc[:, 1:]
    y_header = Y.iloc[:, 0].values
    Y = Y.iloc[:, 1:]
    
    d_mat = pairwise.cosine_distances(X, Y)
    d_mat_df = pd.DataFrame(d_mat, index=x_header)
    d_mat_df.columns = y_header
    
    return d_mat_df

def validate(X, Y, d_mat_df)
    """
    Validate cosine distances function
    X : Bait DataFrame with inst_id as the first column
    Y : All DataFrame with inst_id as the first column
    d_mat_df : result DataFrame from cosine_distances function
    """
    for x_inst in X.iloc[:, 0]:
        x1 = X[X['inst_id'] == x_inst].iloc[:, 1:]
        for y_inst in Y.iloc[:, 0]:
            y1 = Y[Y['inst_id'] == y_inst].iloc[:, 1:]
            dist1 = cosine_distances(x1, y1)[0][0]
            dist2 = d_mat_df[y_inst][x_inst]
            assert abs(dist1 - dist2) < 0.00001, "Something is wrong. {0} is NOT equal with {1}".format(dist1, dist2)