Are Bias Mitigation Techniques for Deep Learning Effective?