Title: Divide and Conquer: Carving Out Concept-based Models out of BlackBox for More Efficient Transfer Learning

Abstract: Building generalizable AI models is one of the primary challenges in the healthcare domain. While radiologists rely on generalizable descriptive rules of abnormality, Neural Networks (NN), often treated as blackboxes, suffer even with a slight shift in input distribution (e.g., scanner type). Fine-tuning a model to transfer knowledge from one domain to another requires a significant amount of labeled data in the target domain. In this paper, we develop a concept-based interpretable model that can be efficiently fine-tuned to an unseen target domain with minimal computational cost. We assume the interpretable component of NN to be approximately domain-invariant.

Concept-based model design either starts with an interpretable by design or a post-hoc-based approach from a Blackbox. Blackbox models are flexible but difficult to explain, while interpretable by-design models are inherently explainable. Yet, interpretable models require extensive machine learning knowledge and tend to be less flexible, potentially underperforming their Blackbox equivalents. My research aims to blur the distinction between a post-hoc explanation of a Blackbox and constructing interpretable models. In the first part of the talk, beginning with a Blackbox, we iteratively carve out a mixture of concept-based interpretable models and a residual network. The interpretable models identify a subset of samples and explain them using First Order Logic (FOL), providing basic reasoning on concepts from the Blackbox. We route the remaining samples through a flexible residual. We repeat the method on the residual network until all the interpretable models explain the desired proportion of data. In the second part of my talk, I will discuss an algorithm to transfer the interpretable models from a source domain to an unseen target domain with minimum training data and computation cost.

Bio: Shantanu Ghosh is a PhD candidate in electrical engineering at Boston University, advised by Prof. Kayhan Batmanghelich. He completed his master's degree in computer science from the University of Florida under the supervision of Dr. Mattia Prosperi. His research interests include exploring various architectures of supervised, unsupervised, and self-supervised deep neural networks to enhance their generalization and interpretability and their applications in medical image analysis. A fun fact: He started his PhD in the ISP program at the University of Pittsburgh. He transferred to BU following his advisor's move.

Title: An interpretable deep learning framework for genome-informed precision oncology 

Abstract: Cancers result from aberrations in cellular signaling systems, typically resulting from driver somatic genome alterations (SGAs) in individual tumors.  Precision oncology requires understanding the cellular state and selecting medications that induce vulnerability in cancer cells under such conditions. To this end, we developed a computational framework consisting of two components: 1) A representation-learning component, which learns a representation of the cellular signaling systems when perturbed by SGAs, using a biologically-motivated and interpretable deep learning model.  2) A drug-response-prediction component, which predicts the response to drugs by leveraging the information of the cellular state of the cancer cells derived by the first component. Our cell-state-oriented framework significantly enhances the accuracy of genome-informed prediction of drug responses in comparison to models that directly use SGAs as inputs. Importantly, our framework enables the prediction of response to chemotherapy agents based on SGAs, thus expanding genome-informed precision oncology beyond molecularly targeted drugs.