Efficient and Scalable Fine-Tune of Language Models for Genome Understanding

Although DNA foundation models have advanced the understanding of genomes, they still face significant challenges in the limited scale and diversity of genomic data. This limitation starkly contrasts with the success of natural language foundation models, which thrive on substantially larger scales. Furthermore, genome understanding involves numerous downstream genome annotation tasks with inherent data heterogeneity, thereby necessitating more efficient and robust fine-tuning methods tailored for genomics. Here, we present \textsc{Lingo}: \textsc{L}anguage prefix f\textsc{In}e-tuning for \textsc{G}en\textsc{O}mes. Unlike DNA foundation models, \textsc{Lingo} strategically leverages natural language foundation models' contextual cues, recalibrating their linguistic knowledge to genomic sequences. \textsc{Lingo} further accommodates numerous, heterogeneous downstream fine-tune tasks by an adaptive rank sampling method that prunes and stochastically reintroduces pruned singular vectors within small computational budgets. Adaptive rank sampling outperformed existing fine-tuning methods on all benchmarked 14 genome understanding tasks, while requiring fewer than 2\% of trainable parameters as genomic-specific adapters. Impressively, applying these adapters on natural language foundation models matched or even exceeded the performance of DNA foundation models. \textsc{Lingo} presents a new paradigm of efficient and scalable genome understanding via genomic-specific adapters on language models.

ProPath: Disease-Specific Protein Language Model for Variant Pathogenicity

Clinical variant classification of pathogenic versus benign genetic variants remains a pivotal challenge in clinical genetics. Recently, the proposition of protein language models has improved the generic variant effect prediction (VEP) accuracy via weakly-supervised or unsupervised training. However, these VEPs are not disease-specific, limiting their adaptation at point-of-care. To address this problem, we propose a disease-specific \textsc{pro}tein language model for variant \textsc{path}ogenicity, termed ProPath, to capture the pseudo-log-likelihood ratio in rare missense variants through a siamese network. We evaluate the performance of ProPath against pre-trained language models, using clinical variant sets in inherited cardiomyopathies and arrhythmias that were not seen during training. Our results demonstrate that ProPath surpasses the pre-trained ESM1b with an over $5\%$ improvement in AUC across both datasets. Furthermore, our model achieved the highest performances across all baselines for both datasets. Thus, our ProPath offers a potent disease-specific variant effect prediction, particularly valuable for disease associations and clinical applicability.

Privacy-Preserving Representation Learning for Text-Attributed Networks with Simplicial Complexes

Although recent network representation learning (NRL) works in text-attributed networks demonstrated superior performance for various graph inference tasks, learning network representations could always raise privacy concerns when nodes represent people or human-related variables. Moreover, standard NRLs that leverage structural information from a graph proceed by first encoding pairwise relationships into learned representations and then analysing its properties. This approach is fundamentally misaligned with problems where the relationships involve multiple points, and topological structure must be encoded beyond pairwise interactions. Fortunately, the machinery of topological data analysis (TDA) and, in particular, simplicial neural networks (SNNs) offer a mathematically rigorous framework to learn higher-order interactions between nodes. It is critical to investigate if the representation outputs from SNNs are more vulnerable compared to regular representation outputs from graph neural networks (GNNs) via pairwise interactions. In my dissertation, I will first study learning the representations with text attributes for simplicial complexes (RT4SC) via SNNs. Then, I will conduct research on two potential attacks on the representation outputs from SNNs: (1) membership inference attack, which infers whether a certain node of a graph is inside the training data of the GNN model; and (2) graph reconstruction attacks, which infer the confidential edges of a text-attributed network. Finally, I will study a privacy-preserving deterministic differentially private alternating direction method of multiplier to learn secure representation outputs from SNNs that capture multi-scale relationships and facilitate the passage from local structure to global invariant features on text-attributed networks.

Measuring the Privacy Leakage via Graph Reconstruction Attacks on Simplicial Neural Networks (Student Abstract)

In this paper, we measure the privacy leakage via studying whether graph representations can be inverted to recover the graph used to generate them via graph reconstruction attack (GRA). We propose a GRA that recovers a graph's adjacency matrix from the representations via a graph decoder that minimizes the reconstruction loss between the partial graph and the reconstructed graph. We study three types of representations that are trained on the graph, i.e., representations output from graph convolutional network (GCN), graph attention network (GAT), and our proposed simplicial neural network (SNN) via a higher-order combinatorial Laplacian. Unlike the first two types of representations that only encode pairwise relationships, the third type of representation, i.e., SNN outputs, encodes higher-order interactions (e.g., homological features) between nodes. We find that the SNN outputs reveal the lowest privacy-preserving ability to defend the GRA, followed by those of GATs and GCNs, which indicates the importance of building more private representations with higher-order node information that could defend the potential threats, such as GRAs.

Human-guided Robot Behavior Learning: A GAN-assisted Preference-based Reinforcement Learning Approach

Human demonstrations can provide trustful samples to train reinforcement learning algorithms for robots to learn complex behaviors in real-world environments. However, obtaining sufficient demonstrations may be impractical because many behaviors are difficult for humans to demonstrate. A more practical approach is to replace human demonstrations by human queries, i.e., preference-based reinforcement learning. One key limitation of the existing algorithms is the need for a significant amount of human queries because a large number of labeled data is needed to train neural networks for the approximation of a continuous, high-dimensional reward function. To reduce and minimize the need for human queries, we propose a new GAN-assisted human preference-based reinforcement learning approach that uses a generative adversarial network (GAN) to actively learn human preferences and then replace the role of human in assigning preferences. The adversarial neural network is simple and only has a binary output, hence requiring much less human queries to train. Moreover, a maximum entropy based reinforcement learning algorithm is designed to shape the loss towards the desired regions or away from the undesired regions. To show the effectiveness of the proposed approach, we present some studies on complex robotic tasks without access to the environment reward in a typical MuJoCo robot locomotion environment. The obtained results show our method can achieve a reduction of about 99.8% human time without performance sacrifice.

Deep Model Compression via Deep Reinforcement Learning

Besides accuracy, the storage of convolutional neural networks (CNN) models is another important factor considering limited hardware resources in practical applications. For example, autonomous driving requires the design of accurate yet fast CNN for low latency in object detection and classification. To fulfill the need, we aim at obtaining CNN models with both high testing accuracy and small size/storage to address resource constraints in many embedded systems. In particular, this paper focuses on proposing a generic reinforcement learning based model compression approach in a two-stage compression pipeline: pruning and quantization. The first stage of compression, i.e., pruning, is achieved via exploiting deep reinforcement learning (DRL) to co-learn the accuracy of CNN models updated after layer-wise channel pruning on a testing dataset and the FLOPs, number of floating point operations in each layer, updated after kernel-wise variational pruning using information dropout. Layer-wise channel pruning is to remove unimportant kernels from the input channel dimension while kernel-wise variational pruning is to remove unimportant kernels from the 2D-kernel dimensions, namely, height and width. The second stage, i.e., quantization, is achieved via a similar DRL approach but focuses on obtaining the optimal weight bits for individual layers. We further conduct experimental results on CIFAR-10 and ImageNet datasets. For the CIFAR-10 dataset, the proposed method can reduce the size of VGGNet by 9x from 20.04MB to 2.2MB with 0.2% accuracy increase. For the ImageNet dataset, the proposed method can reduce the size of VGG-16 by 33x from 138MB to 4.14MB with no accuracy loss.

Relationship Explainable Multi-objective Optimization Via Vector Value Function Based Reinforcement Learning

Solving multi-objective optimization problems is important in various applications where users are interested in obtaining optimal policies subject to multiple, yet often conflicting objectives. A typical approach to obtain optimal policies is to first construct a loss function that is based on the scalarization of individual objectives, and then find the optimal policy that minimizes the loss. However, optimizing the scalarized (and weighted) loss does not necessarily provide a guarantee of high performance on each possibly conflicting objective. In this paper, we propose a vector value based reinforcement learning approach that seeks to explicitly learn the inter-objective relationship and optimize multiple objectives based on the learned relationship. In particular, the proposed method is to first define relationship matrix, a mathematical representation of the inter-objective relationship, and then create one actor and multiple critics that can co-learn the relationship matrix and action selection. The proposed approach can quantify the inter-objective relationship via reinforcement learning when the impact of one objective on another is unknown a prior. We also provide rigorous convergence analysis of the proposed approach and present a quantitative evaluation of the approach based on two testing scenarios.

Relationship Explainable Multi-objective Reinforcement Learning with Semantic Explainability Generation

Solving multi-objective optimization problems is important in various applications where users are interested in obtaining optimal policies subject to multiple, yet often conflicting objectives. A typical approach to obtain optimal policies is to first construct a loss function that is based on the scalarization of individual objectives, and then find the optimal policy that minimizes the loss. However, optimizing the scalarized (and weighted) loss does not necessarily provide guarantee of high performance on each possibly conflicting objective because it is challenging to assign the right weights without knowing the relationship among these objectives. Moreover, the effectiveness of these gradient descent algorithms is limited by the agent's ability to explain their decisions and actions to human users. The purpose of this study is two-fold. First, we propose a vector value function based multi-objective reinforcement learning (V2f-MORL) approach that seeks to quantify the inter-objective relationship via reinforcement learning (RL) when the impact of one objective on others is unknown a prior. In particular, we construct one actor and multiple critics that can co-learn the policy and inter-objective relationship matrix (IORM), quantifying the impact of objectives on each other, in an iterative way. Second, we provide a semantic representation that can uncover the trade-off of decision policies made by users to reconcile conflicting objectives based on the proposed V2f-MORL approach for the explainability of the generated behaviors subject to given optimization objectives. We demonstrate the effectiveness of the proposed approach via a MuJoCo based robotics case study.