Píetro Lió, Martin J. Bishop · Methods in molecular biology · 2008
DNA and amino acid sequences contain information about both the phylogenetic relationships among species and the evolutionary processes that caused the sequences to divergence. Mathematical and statistical methods try to detect this information to determine how and why DNA and protein molecules work the way they do. This chapter describes some of the models of evolution of biological sequences most widely used. It first focuses on single nucleotide/amino acid replacement rate models. Then it discusses the modelling of evolution at gene and protein module levels. The chapter concludes with speculations about the future use of molecular evolution studies using genomic and proteomic data.
Cited by 9
Jan Niehues, Ngoc-Quan Pham · 2019
Recently, significant improvements have been achieved in various natural language processing tasks using neural sequence-to-sequence models. While aiming for the best generation quality is important, ultimately it is also necessary to develop models that can assess the quality of their output.
Cited by 0Open access
Lin Geng Foo, Tianjiao Li, Hossein Rahmani et al. · Monash University Research Portal (Monash University) · 2023
We propose a Unified Pose Sequence Modeling approach to unify heterogeneous human behavior understanding tasks based on pose data, e.g., action recognition, 3D pose estimation and 3D early action prediction. A major obstacle is that different pose-based tasks require different output data formats. Specifically, the action recognition and prediction tasks require class predictions as outputs, while 3D pose estimation requires a human pose output, which limits existing methods to leverage task-specific network architectures for each task. Hence, in this paper, we propose a novel Unified Pose Sequence (UPS) model to unify heterogeneous output formats for the aforementioned tasks by considering text-based action labels and coordinate-based human poses as language sequences. Then, by optimizing a single auto-regressive transformer, we can obtain a unified output sequence that can handle all the aforementioned tasks. Moreover, to avoid the interference brought by the heterogeneity between different tasks, a dynamic routing mechanism is also proposed to empower our UPS with the ability to learn which subsets of parameters should be shared among different tasks. To evaluate the efficacy of the proposed UPS, extensive experiments are conducted on four different tasks with four popular behavior understanding benchmarks.
Cited by 40Open access
Anna Bernasconi · Lecture notes in business information processing · 2023
Cited by 0
Kyungwoo Song · 2020
Context modeling helps understand the data, such as sentence or user behavior. Contextual information captures the important underlying feature, and it enhances the relationship between data instances or hidden representations. As the importance of the sequential model grows, so does the importance of the sequential contextual modeling. Under the sequential data, we need to consider the context change over time. In this paper, we present our research works on context modeling and its dynamics modeling over time. Furthermore, we extend our research to handle the multi-granularity of sequential context modeling to consider rich context representations.
Cited by 0Open access
Rahmani, Hossein · Lancaster EPrints (Lancaster University) · 2023
Cited by 0Open access
Gaya Prasad Yadav (198109), Sonal Shree (676584), Ruchi Maurya (676585) et al. · Figshare · 2015
<p>(<b>A</b>) <b>Sequence alignment of MtSerB2</b> with sequences of Phosphoserine phosphatases from <i>M. avium</i> (MavSerB), <i>P. gingivalis</i> (PgSerB653), <i>P. gingivalis</i> (PgSerB1170), <i>M. tuberculosis</i> (MtSerB1), MjPSPase (<i>M. janaschii</i>) and HsPSPase <i>(Homo sapiens)</i>. Three conserved motifs of the PSP domain are shown in <i>red</i>. The ACT1 and ACT2 domains are colored <i>blue</i> and <i>green</i> respectively Secondary structural elements are also indicated. The sequences highlighted in <i>red</i> represents high consensus whereas those in <i>blue</i> represents low consensus (<b>B</b>) <b>Modeled structure of MtSerB2</b>. MtSerB2 structure was modeled using <i>M. avium</i> SerB structure (PDB: 3P96) and Modeler 9.10. The monomeric and dimeric associations are depicted and the individual domains are labelled. Gly residues important for binding ligands in ACT domains are shown in ‘stick’ representation and labeled for clarity. Selected catalytic residues on the PSP domain are also labeled and depicted.</p>
Cited by 0Open access
Aishwarya Asesh · 2022
A lot of research has been done on music generation. Previous attempts at creating music have been successful at reproducing training tracks using a variety of techniques involving time inference models. This research demonstrates that first-order and k-order Markov models can be used to both reproduce and synthesize the music from vintage 8-bit video games, one of the very rare combination of music synthesis using 8-bit temporal inference techniques. The study highlights ways for generating infinite sequential (complex bi-directional) monotones and polytones, on demand along with context structuring. Using proposed approach, Ising model and Metropolis algorithm, any application of sequence modeling can benefit, like marketeers can be provided with the most probable Next-Best-Action to recommend the end user, usually that is not accessible by any conventional sequence models.
Cited by 0
Simran Arora, Christopher Ré, Stanford University · Stanford Digital Repository · 2026
Cited by 0Open access
Rakshitha Bharadwaj, Vaishnavi Bhat, Yaswanth Raj et al. · International Research Journal of Modernization in Engineering Technology and Science · 2024
In AIML, transformation and sequence modeling represent two distinctive approaches to processing user input.Transformation revolves around rule-based manipulations of textual data, where predefined patterns trigger specific responses.This method, ideal for creating conversational interfaces, relies on explicit rules for immediate interaction.In contrast, sequence modeling in AIML adopts advanced neural network architectures like RNNs, LSTMs, or Transformers to understand and generate responses to sequential data, capturing nuanced dependencies and context.While transformation excels in straightforward scenarios with predefined rules, sequence modeling accommodates dynamic and context-dependent interactions.This duality underscores AIML's versatility, employing rule-based transformations for simplicity and neural network-driven sequence modeling for more sophisticated, context-aware conversational AI applications.Understanding these complementary paradigms provides a nuanced perspective on AIML's diverse strategies for processing and responding to user input.In AIML, transformation employs rule-based patterns for immediate responses, suitable for well-defined interactions.On the other hand, sequence modeling utilizes advanced neural networks to comprehend nuanced dependencies in sequential data, excelling in complex and context-aware scenarios.These contrasting methodologies underscore AIML's adaptability, catering to both straightforward rule-based conversations and dynamic, context-dependent interactions.
Cited by 0Open access
Awni Hannun · Distill · 2017
Cited by 73
Emma Jo Ciccarelli (18831406), Zachary Wing (18831409), Moshe Bendelstein (18831412) et al. · Figshare · 2024
<p>(A) Annotated TIG-2 and TIG-3 isoform B amino acid sequences. The signal sequence (black) is underlined, and the cleavage site is in boldface. The prodomain (blue-violet in TIG-2, green-cyan in TIG-3) includes the cleavage site, which separates the prodomain from the mature ligand (dark khaki in TIG-2, gray-beige in TIG-3). Cysteine residues participating in intrachain disulfide bonds are highlighted in gold. The lysine residue that replaces the dimerization cysteine is colored yellow-green. (B) Subfamily, species, Uniprot entry, signal sequence, and cleavage site for modeled TGF-β family members. (a) [<a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1011324#pgen.1011324.ref060" target="_blank">60</a>] (b) SignalP 6.0. (c) Manual sequence analysis. (d) [<a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1011324#pgen.1011324.ref052" target="_blank">52</a>].</p> <p>(EPS)</p>
Cited by 0Open access
Zhang, Jing-Xuan, Ling, Zhen-Hua, Liu, Li-Juan et al. · arXiv (Cornell University) · 2018
In this paper, a neural network named Sequence-to-sequence ConvErsion NeTwork (SCENT) is presented for acoustic modeling in voice conversion. At training stage, a SCENT model is estimated by aligning the feature sequences of source and target speakers implicitly using attention mechanism. At conversion stage, acoustic features and durations of source utterances are converted simultaneously using the unified acoustic model. Mel-scale spectrograms are adopted as acoustic features which contain both excitation and vocal tract descriptions of speech signals. The bottleneck features extracted from source speech using an automatic speech recognition (ASR) model are appended as auxiliary input. A WaveNet vocoder conditioned on Mel-spectrograms is built to reconstruct waveforms from the outputs of the SCENT model. It is worth noting that our proposed method can achieve appropriate duration conversion which is difficult in conventional methods. Experimental results show that our proposed method obtained better objective and subjective performance than the baseline methods using Gaussian mixture models (GMM) and deep neural networks (DNN) as acoustic models. This proposed method also outperformed our previous work which achieved the top rank in Voice Conversion Challenge 2018. Ablation tests further confirmed the effectiveness of several components in our proposed method.
Cited by 0Open access
Aynaz Taheri, Kevin Gimpel, Tanya Berger‐Wolf · Applied Network Science · 2019
We propose sequence-to-sequence architectures for graph representation learning in both supervised and unsupervised regimes. Our methods use recurrent neural networks to encode and decode information from graph-structured data. Recurrent neural networks require sequences, so we choose several methods of traversing graphs using different types of substructures with various levels of granularity to generate sequences of nodes for encoding. Our unsupervised approaches leverage long short-term memory (LSTM) encoder-decoder models to embed the graph sequences into a continuous vector space. We then represent a graph by aggregating its graph sequence representations. Our supervised architecture uses an attention mechanism to collect information from the neighborhood of a sequence. The attention module enriches our model in order to focus on the subgraphs that are crucial for the purpose of a graph classification task. We demonstrate the effectiveness of our approaches by showing improvements over the existing state-of-the-art approaches on several graph classification tasks.
Cited by 6Open access
Gary Ren, Xiaochuan Ni, Manish Malik et al. · 2018
Understanding conversations is crucial to enabling conversational search in technologies such as chatbots, digital assistants, and smart home devices that are becoming increasingly popular. Conventional search engines are powerful at answering open domain queries but are mostly capable of stateless search. In this paper, we define a conversational query as a query that depends on the context of the current conversation, and we formulate the conversational query understanding problem as context-aware query reformulation, where the goal is to reformulate the conversational query into a search engine friendly query in order to satisfy users» information needs in conversational settings. Such context-aware query reformulation problem lends itself to sequence to sequence modeling. We present a large scale open domain dataset of conversational queries and various sequence to sequence models that are learned from this dataset. The best model correctly reformulates over half of all conversational queries, showing the potential of sequence to sequence modeling for this task.
Cited by 54Open access
Patricia Sardinha Leonardo, Rodrigo Leal de Paiva Carvalho, Sérgio Gomes da Silva et al. · LA Referencia (Red Federada de Repositorios Institucionales de Publicaciones Científicas) · 2024
Cited by 0Open access
Michael Gribskov · 2007
Homology-based modeling, or more simply homology modeling, is becoming increasingly important. Models accurate to within 1-2 Å RMS deviation from the actual three-dimensional structure can be produced by homology modeling in favorable cases. With the number of known threedimensional structures projected to grow to more than ten thousand structures by the end of the decade, it will be possible to model a significant fraction of all sequences using these approaches. In homology modeling, a known three-dimensional structure is used as a template from which a three-dimensional model of a novel sequence is calculated. In the first stages of this procedure, the critical steps are the location of an appropriate homologous structure to use as the template, and the correct mapping of the novel sequence onto the known structure. The detection of a homologous structure involves the inference that the structure is indeed homologous based on a sequence or structure-based comparison. This inference ...
Cited by 0
Vasileios Lioutas · 2020
Sequence modeling is one of the fundamental tasks in machine learning. As such, it has been studied for years and various different approaches have been proposed. Today, non-autoregressive attention-based methods are deemed standard for modeling sequences. Methods like Transformer which use self-attention have been shown to be capable of producing state-of-the-art results. The major drawback of self-attention is that it requires a quadratic time complexity O(n 2 ) for the processing of a sequence. This makes the use of self-attention expensive and slow when it comes to long sequences. In this thesis, we aim to reduce this complexity to linear time without using attention. To do so, we introduce Time-aware Large Kernel (TaLK) Convolutions, a novel adaptive convolution operation that learns to predict the size of a summation kernel instead of using a fixed-sized kernel matrix. This method has a time complexity of O(n). We evaluated our method using standard benchmark datasets in four different NLP tasks: machine translation, language modeling, abstractive text summarization and text classification. We show that our proposed adaptive convolution method is capable of achieving state-of-the-art results without using attention and with a significantly faster execution time and smaller memory footprint.
Cited by 0Open access
Martinez, Matthew Thomas · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 2019
Abstract not provided.
Cited by 0Open access
K. T. Deepak · IEEE RESOURCE CENTERS · 2024
Cited by 0Open access