More Like this

1. MULTI-SCALE DENSE NETWORKS FOR RESOURCE EFFICIENT IMAGE CLASSIFICATION

   Gao Huang, Danlu Chen (April 2018, ICLR 2018)

   In this paper we investigate image classification with computational resource lim- its at test time. Two such settings are: 1. anytime classification, where the net- work’s prediction for a test example is progressively updated, facilitating the out- put of a prediction at any time; and 2. budgeted batch classification, where a fixed amount of computation is available to classify a set of examples that can be spent unevenly across “easier” and “harder” inputs. In contrast to most prior work, such as the popular Viola and Jones algorithm, our approach is based on convolutional neural networks. We train multiple classifiers with varying resource demands, which we adaptively apply during test time. To maximally re-use computation between the classifiers, we incorporate them as early-exits into a single deep convolutional neural network and inter-connect them with dense connectivity. To facilitate high quality classification early on, we use a two-dimensional multi-scale network architecture that maintains coarse and fine level features all-throughout the network. Experiments on three image-classification tasks demonstrate that our framework substantially improves the existing state-of-the-art in both settings. 

   more » « less
2. Predicting transcriptional responses to cold stress across plant species

   https://doi.org/10.1073/pnas.2026330118  

   Meng, Xiaoxi; Liang, Zhikai; Dai, Xiuru; Zhang, Yang; Mahboub, Samira; Ngu, Daniel W.; Roston, Rebecca L.; Schnable, James C. (March 2021, Proceedings of the National Academy of Sciences)

   Although genome-sequence assemblies are available for a growing number of plant species, gene-expression responses to stimuli have been cataloged for only a subset of these species. Many genes show altered transcription patterns in response to abiotic stresses. However, orthologous genes in related species often exhibit different responses to a given stress. Accordingly, data on the regulation of gene expression in one species are not reliable predictors of orthologous gene responses in a related species. Here, we trained a supervised classification model to identify genes that transcriptionally respond to cold stress. A model trained with only features calculated directly from genome assemblies exhibited only modest decreases in performance relative to models trained by using genomic, chromatin, and evolution/diversity features. Models trained with data from one species successfully predicted which genes would respond to cold stress in other related species. Cross-species predictions remained accurate when training was performed in cold-sensitive species and predictions were performed in cold-tolerant species and vice versa. Models trained with data on gene expression in multiple species provided at least equivalent performance to models trained and tested in a single species and outperformed single-species models in cross-species prediction. These results suggest that classifiers trained on stress data from well-studied species may suffice for predicting gene-expression patterns in related, less-studied species with sequenced genomes. 

   more » « less
3. Adversarial Attack on Graph Structured Data

   Dai, Hanjun; Li, Hui; Tian, Tian; Huang, Xin; Wang, Lin; Zhu, Jun; Song, Le (July 2018, Proceedings of the 35th International Conference on Machine Learning, PMLR)

   Deep learning on graph structures has shown exciting results in various applications. However, few attentions have been paid to the robustness of such models, in contrast to numerous research work for image or text adversarial attack and defense. In this paper, we focus on the adversarial attacks that fool deep learning models by modifying the combinatorial structure of data. We first propose a reinforcement learning based attack method that learns the generalizable attack policy, while only requiring prediction labels from the target classifier. We further propose attack methods based on genetic algorithms and gradient descent in the scenario where additional prediction confidence or gradients are available. We use both synthetic and real-world data to show that, a family of Graph Neural Network models are vulnerable to these attacks, in both graph-level and node-level classification tasks. We also show such attacks can be used to diagnose the learned classifiers. 

   more » « less
4. Transformer for Gene Expression Modeling (T-GEM): An Interpretable Deep Learning Model for Gene Expression-Based Phenotype Predictions

   https://doi.org/10.3390/cancers14194763  

   Zhang, Ting-He; Hasib, Md Musaddaqul; Chiu, Yu-Chiao; Han, Zhi-Feng; Jin, Yu-Fang; Flores, Mario; Chen, Yidong; Huang, Yufei (October 2022, Cancers)

   Deep learning has been applied in precision oncology to address a variety of gene expression-based phenotype predictions. However, gene expression data’s unique characteristics challenge the computer vision-inspired design of popular Deep Learning (DL) models such as Convolutional Neural Network (CNN) and ask for the need to develop interpretable DL models tailored for transcriptomics study. To address the current challenges in developing an interpretable DL model for modeling gene expression data, we propose a novel interpretable deep learning architecture called T-GEM, or Transformer for Gene Expression Modeling. We provided the detailed T-GEM model for modeling gene–gene interactions and demonstrated its utility for gene expression-based predictions of cancer-related phenotypes, including cancer type prediction and immune cell type classification. We carefully analyzed the learning mechanism of T-GEM and showed that the first layer has broader attention while higher layers focus more on phenotype-related genes. We also showed that T-GEM’s self-attention could capture important biological functions associated with the predicted phenotypes. We further devised a method to extract the regulatory network that T-GEM learns by exploiting the attributions of self-attention weights for classifications and showed that the network hub genes were likely markers for the predicted phenotypes. 

   more » « less
5. Combination of deep neural network with attention mechanism enhances the explainability of protein contact prediction

   https://doi.org/10.1002/prot.26052  

   Chen, Chen; Wu, Tianqi; Guo, Zhiye; Cheng, Jianlin (February 2021, Proteins: Structure, Function, and Bioinformatics)

   Abstract Deep learning has emerged as a revolutionary technology for protein residue‐residue contact prediction since the 2012 CASP10 competition. Considerable advancements in the predictive power of the deep learning‐based contact predictions have been achieved since then. However, little effort has been put into interpreting the black‐box deep learning methods. Algorithms that can interpret the relationship between predicted contact maps and the internal mechanism of the deep learning architectures are needed to explore the essential components of contact inference and improve their explainability. In this study, we present an attention‐based convolutional neural network for protein contact prediction, which consists of two attention mechanism‐based modules: sequence attention and regional attention. Our benchmark results on the CASP13 free‐modeling targets demonstrate that the two attention modules added on top of existing typical deep learning models exhibit a complementary effect that contributes to prediction improvements. More importantly, the inclusion of the attention mechanism provides interpretable patterns that contain useful insights into the key fold‐determining residues in proteins. We expect the attention‐based model can provide a reliable and practically interpretable technique that helps break the current bottlenecks in explaining deep neural networks for contact prediction. The source code of our method is available athttps://github.com/jianlin-cheng/InterpretContactMap. 

   more » « less