arXiv subjects

Farshad Noravesh

Publications and source records attributed to Farshad Noravesh.

14 recordsLinked to original sources

arXiv2026

VQ-SAD: Vector Quantized Structure Aware Diffusion For Molecule Generation

Many diffusion based molecule generation methods ignore the symbolic information of molecules and represent the atom and bond type as one hot representation. Methods based on Morgan fingerprints produce hash collisions and are hard to embed into a continuous space without information loss and random fingerprints correspond to no valid molecule. To circumvent this issue we use another paradigm and consider atom and bond codes as latent variables of VQ-VAE. We introduce VQ-SAD which first trains a VQ-VAE and uses the frozen pretrained VQ-VAE model and considers the codebooks for both atom and bond types as tokenizers for the downstream diffusion process. VQ-SAD is a neuro-symbolic model that utilizes both symbolic and neural structural information for a diffusion based model with learnable forward process. The large discrete code space provides a more balanced atom and bond types which enhances the denoising process. VQ-VAE slightly outperforms SOTA models for diffusion based molecule generation on QM9 and ZINC250k datasets.

cs.LG↗

arXiv2025

Are Graph Attention Networks Able to Model Structural Information?

Graph Attention Networks (GATs) have emerged as powerful models for learning expressive representations from such data by adaptively weighting neighboring nodes through attention mechanisms. However, most existing approaches primarily rely on node attributes and direct neighborhood connections, often overlooking rich structural patterns that capture higher-order topological information crucial for many real-world datasets. In this work, we present the Graph Structure Attention Network (GSAT), a novel extension of GAT that jointly integrates attribute-based and structure-based representations for more effective graph learning. GSAT incorporates structural features derived from anonymous random walks (ARWs) and graph kernels to encode local topological information, enabling attention mechanisms to adapt based on the underlying graph structure. This design enhances the model's ability to discern meaningful relational dependencies within complex data. Comprehensive experiments on standard graph classification and regression benchmarks demonstrate that GSAT achieves consistent improvements over state-of-the-art graph learning methods, highlighting the value of incorporating structural context for representation learning on graphs.

cs.LG↗

arXiv2025

LGRPool: Hierarchical Graph Pooling Via Local-Global Regularisation

Hierarchical graph pooling(HGP) are designed to consider the fact that conventional graph neural networks(GNN) are inherently flat and are also not multiscale. However, most HGP methods suffer not only from lack of considering global topology of the graph and focusing on the feature learning aspect, but also they do not align local and global features since graphs should inherently be analyzed in a multiscale way. LGRPool is proposed in the present paper as a HGP in the framework of expectation maximization in machine learning that aligns local and global aspects of message passing with each other using a regularizer to force the global topological information to be inline with the local message passing at different scales through the representations at different layers of HGP. Experimental results on some graph classification benchmarks show that it slightly outperforms some baselines.

cs.LG↗

arXiv2025

Attending To Syntactic Information In Biomedical Event Extraction Via Graph Neural Networks

Many models are proposed in the literature on biomedical event extraction(BEE). Some of them use the shortest dependency path(SDP) information to represent the argument classification task. There is an issue with this representation since even missing one word from the dependency parsing graph may totally change the final prediction. To this end, the full adjacency matrix of the dependency graph is used to embed individual tokens using a graph convolutional network(GCN). An ablation study is also done to show the effect of the dependency graph on the overall performance. The results show a significant improvement when dependency graph information is used. The proposed model slightly outperforms state-of-the-art models on BEE over different datasets.

cs.CL↗

arXiv2024

BioNCERE: Non-Contrastive Enhancement For Relation Extraction In Biomedical Texts

State-of-the-art models for relation extraction (RE) in the biomedical domain consider finetuning BioBERT using classification, but they may suffer from the anisotropy problem. Contrastive learning methods can reduce this anisotropy phenomena, and also help to avoid class collapse in any classification problem. In the present paper, a new training method called biological non-contrastive relation extraction (BioNCERE) is introduced for relation extraction without using any named entity labels for training to reduce annotation costs. BioNCERE uses transfer learning and non-contrastive learning to avoid full or dimensional collapse as well as bypass overfitting. It resolves RE in three stages by leveraging transfer learning two times. By freezing the weights learned in previous stages in the proposed pipeline and by leveraging non-contrastive learning in the second stage, the model predicts relations without any knowledge of named entities. Experiments have been done on SemMedDB that are almost similar to State-of-the-art performance on RE without using the information of named entities.

cs.CL↗

arXiv2023

Transferring Neural Potentials For High Order Dependency Parsing

High order dependency parsing leverages high order features such as siblings or grandchildren to improve state of the art accuracy of current first order dependency parsers. The present paper uses biaffine scores to provide an estimate of the arc scores and is then propagated into a graphical model. The inference inside the graphical model is solved using dual decomposition. The present algorithm propagates biaffine neural scores to the graphical model and by leveraging dual decomposition inference, the overall circuit is trained end-to-end to transfer first order informations to the high order informations.

cs.CL↗

arXiv2023

Semantic Tagging with LSTM-CRF

In the present paper, two models are presented namely LSTM-CRF and BERT-LSTM-CRF for semantic tagging of universal semantic tag dataset. The experiments show that the first model is much easier to converge while the second model that leverages BERT embedding, takes a long time to converge and needs a big dataset for semtagging to be effective.

cs.CL↗

arXiv2023

Semantic Operator Prediction and Applications

In the present paper, semantic parsing challenges are briefly introduced and QDMR formalism in semantic parsing is implemented using sequence to sequence model with attention but uses only part of speech(POS) as a representation of words of a sentence to make the training as simple and as fast as possible and also avoiding curse of dimensionality as well as overfitting. It is shown how semantic operator prediction could be augmented with other models like the CopyNet model or the recursive neural net model.

cs.CL↗

arXiv2022

Identifying OCRs in cfDNA WGS Data by Correlation Clustering

In the recent decade, the emergence of liquid biopsy has significantly improved cancer monitoring and detection. Dying cells, including those originating from tumors, shed their DNA into the bloodstream and contribute to a pool of circulating fragments called cell-free DNA (cfDNA). Identifying the tissue origin of these DNA fragments from their epigenetic features has implications in various clinical contexts. Open chromatin regions (OCRs) are important epigenetic features of DNA that reflect cell types of origin. Profiling these features by DNase-seq, ATAC-seq, and histone ChIP-seq provides insights into tissue-specific and disease-specific regulatory mechanisms. Integration of genomic and epigenomic features for cancer detection by liquid biopsy has previously been reported. However, many multimodal analyses require large amounts of cfDNA input and/or multiple types of experiments to cover the genomic and epigenomic aspects of a single sample which is cost and time prohibitive. Thus, methods that capture genomic and epigenomic profiles in a single experiment type with low input requirements are of importance. Predicting OCRs from whole genome sequencing (WGS) data is one such approach. Here, we applied a correlation clustering algorithm to predict OCRs. We used local sequencing depth as input to our algorithm. Multiple processing steps were then applied as follows: count normalization, discrete Fourier transform conversion, graph construction, graph cut optimization by linear programming, and clustering. To validate the proposed method, we compared the output of our predictions (OCR vs. non-OCR) with previously validated open chromatin regions related to human blood samples of the ATAC-db. The percentage of overlap between them is greater than 67%.

q-bio.GN↗

arXiv2022

A Bayesian Approach To Graph Partitioning

A new algorithm based on bayesian inference for learning local graph conductance based on Gaussian Process(GP) is given that uses advanced MCMC convergence ideas to create a scalable and fast algorithm for convergence to stationary distribution which is provided to learn the bahavior of conductance when traversing the indirected weighted graph. First metric embedding is used to represent the vertices of the graph. Then, uniform induced conductance is calculated for training points. Finally, in the learning step, a gaussian process is used to approximate the uniform induced conductance. MCMC is used to measure uncertainty of estimated hyper-parameters.

cs.LG↗

arXiv2022

Finding Many Sparse Cuts Using Entropy Maximization

A randomized algorithm for finding sparse cuts is given which is based on constructing a dual markov chain called multiscale rings process(MRP) and a new concept of entropy. It is shown how the time to absorption of the dual process measures the connectedness of the graph and mixing of the corresponding markov process which is then utilized to do clustering. The second algorithm uses the entropy which provides a new methodology and a set of tools to think about sparse cuts as well as sparsification of a graph.

math.PR↗

arXiv2022

Sparse Non-Convex Optimization For Higher Moment Portfolio Management

One of the reasons that higher order moment portfolio optimization methods are not fully used by practitioners in investment decisions is the complexity that these higher moments create by making the optimization problem nonconvex. Many few methods and theoretical results exists in the literature, but the present paper uses the method of successive convex approximation for the mean-variance-skewness problem.

cs.CE↗

arXiv2022

Some connections between higher moments portfolio optimization methods

In this paper, different approaches to portfolio optimization having higher moments such as skewness and kurtosis are classified so that the reader can observe different paradigms and approaches in this field of research which is essential for practitioners in Hedge Funds in particular. Several methods based on different paradigms such as utility approach and multi-objective optimization are reviewed and the advantage and disadvantageous of these ideas are explained. Keywords: multi-objective optimization, portfolio optimization, scalarization, utility

cs.CE↗

arXiv2021

Exact Post-selection Inference For Tracking S&P500

The problem that is solved in this paper is known as index tracking. The method of Lasso is used to reduce the dimensions of S&P500 index which has many applications in both investment and portfolio management algorithms. The novelty of this paper is that post-selection inference is used to have better modeling and inference for Lasso approach to index tracking. Both confidence intervals and curves indicate that the performance of Lasso type method for dimension reduction of S&P500 is remarkably high. Keywords: index tracking, lasso, post-selection inference, S&P500

q-fin.ST↗