1. Towards Trustworthy AI

1. Are DNNs safe and trustworthy? [Robust ML]
2. Why do DNNs perform so well? [Interpretable ML]

We focus on Interpretability. Interpretability is crucial to enable Trustworthy AI.

Accuracy & Interpretability trade-offs exists.

1. Interpretability tool

Enable Trust, Safety & Transparency.

2. Interpretable DNN by design

3. Evaluation

References

1. Lily Weng's Talk