MADA-Unet: MultiRes-scale attentive dense-aggregating Unet for worst-case transient IR-drop prediction

Key Points

• Worst-case transient IR-drop predictions showed significant accuracy improvements using MADA-Unet.
• The model predicted IR-drop variations with a notable error reduction of 15% across tested scenarios.
• MADA-Unet employs a multi-resolution architecture combined with attention mechanisms for efficient learning.
• This approach highlights the need for advanced modeling techniques in power integrity assessments.