Constrained Graph Variational Autoencoders for Molecule Design
- Qi Liu, Miltiadis Allamanis, Marc Brockschmidt, Alexander L. Gaunt
- Computer ScienceNeural Information Processing Systems
- 23 May 2018
A variational autoencoder model in which both encoder and decoder are graph-structured is proposed and it is shown that by using appropriate shaping of the latent space, this model allows us to design molecules that are (locally) optimal in desired properties.
Improving gene set analysis of microarray data by SAM-GS
It is concluded that GSEA has important limitations as a gene-set analysis approach for microarray experiments for identifying biological pathways associated with a binary phenotype and an alternative statistically-sound method is proposed, SAM-GS.
Sentence-State LSTM for Text Representation
This work investigates an alternative LSTM structure for encoding text, which consists of a parallel state for each word, and shows that the proposed model has strong representation power, giving highly competitive performances compared to stacked BiLSTM models with similar parameter numbers.
HIT: linking herbal active ingredients to targets
A comprehensive and fully curated database for Herb Ingredients’ Targets (HIT) has been constructed to complement above resources and contains 5208 entries about 1301 known protein targets affected by 586 herbal compounds from more than 1300 reputable Chinese herbs.
Transfer Learning on Heterogenous Feature Spaces via Spectral Transformation
- Xiaoxiao Shi, Qi Liu, W. Fan, Philip S. Yu, Ruixin Zhu
- Computer ScienceIEEE International Conference on Data Mining
- 13 December 2010
A Bayesian-based approach is applied to model the relationship between different output spaces and extracted examples from heterogeneous sources can reduce the error rate by as much as~50\%, compared with the methods using only the examples from the target task.
Feature Distillation: DNN-Oriented JPEG Compression Against Adversarial Examples
- Zihao Liu, Qi Liu, Tao Liu, Yanzhi Wang, Wujie Wen
- Computer ScienceComputer Vision and Pattern Recognition
- 14 March 2018
The experimental results show that proposed “feature distillation” can significantly surpass the latest input-transformation based mitigations such as Quilting and TV Minimization in three aspects, including defense efficiency, accuracy of benign images after defense, and processing time per image.
Quaternion Knowledge Graph Embedding
This paper moves beyond standard complex representations, adopting expressive hypercomplex representations for learning representations of entities and relations, and demonstrates that QuatE achieves state-of-the-art performance on four well-established knowledge graph completion benchmarks.
An improved method to construct basic probability assignment based on the confusion matrix for classification problem
Insertion-based Decoding with Automatically Inferred Generation Order
- Jiatao Gu, Qi Liu, Kyunghyun Cho
- Computer ScienceTransactions of the Association for Computational…
- 4 February 2019
This work proposes a novel decoding algorithm— InDIGO—which supports flexible sequence generation in arbitrary orders through insertion operations, and extends Transformer, a state-of-the-art sequence generation model, to efficiently implement this approach.