Claim Missing Document
Check
Articles

Found 8 Documents
Search

Reducing Adversarial Vulnerability through Adaptive Training Batch Size Ken Sasongko; Adila Alfa Krisnadhi; Mohamad Ivan Fanany
Jurnal Ilmu Komputer dan Informasi Vol 14, No 1 (2021): Jurnal Ilmu Komputer dan Informasi (Journal of Computer Science and Information
Publisher : Faculty of Computer Science - Universitas Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21609/jiki.v14i1.907

Abstract

Neural networks possess an ability to generalize well to data distribution, to an extent that they are capable of fitting to a randomly labeled data. But they are also known to be extremely sensitive to adversarial examples. Batch Normalization (BatchNorm), very commonly part of deep learning architecture, has been found to increase adversarial vulnerability. Fixup Initialization (Fixup Init) has been shown as an alternative to BatchNorm, which can considerably strengthen the networks against adversarial examples. This robustness can be improved further by employing smaller batch size in training. The latter, however, comes with a tradeoff in the form of a significant increase of training time (up to ten times longer when reducing batch size from the default 128 to 8 for ResNet-56). In this paper, we propose a workaround to this problem by starting the training with a small batch size and gradually increase it to larger ones during training. We empirically show that our proposal can still improve adversarial robustness (up to 5.73\%) of ResNet-56 with Fixup Init and default batch size of 128. At the same time, our proposal keeps the training time considerably shorter (only 4 times longer, instead of 10 times).
LexID: The Metadata and Semantic Knowledge Graph Construction of Indonesian Legal Document Nur Siti Muninggar; Adila Alfa Krisnadhi
Jurnal Ilmu Komputer dan Informasi Vol. 16 No. 1 (2023): Jurnal Ilmu Komputer dan Informasi (Journal of Computer Science and Informatio
Publisher : Faculty of Computer Science - Universitas Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21609/jiki.v16i1.1096

Abstract

The Legal Fiction principle stipulates that the government needs to ensure the public availability of all of their legal documents. Unfortunately, the text-based search services they provide cannot return satisfactory answers in retrieval scenarios requiring proper representation of relationships between various legal documents. A key problem here is the lack of explicit representation of such relationships behind the employed retrieval engines. We aim to address this problem by proposing LexID knowledge graph (KG) that provides an explicit knowledge representation for Indonesian legal domain usable for such retrieval purposes. The KG contains both legal metadata information and semantic content of the legal clauses of the legal document's articles, modeled using formal vocabulary from the LexID ontology also presented in this paper. The KG is constructed from thousands of Indonesian legal documents. Since the procedure of writing a legal document regulated by the government is clear and detailed, we use a rule-based approach to construct our KG. At the end, we describe several use cases of the KG to address different retrieval needs. In Addition, we evaluated the quality of our KG by measuring its ability to answer questions and got that LexID can answer questions with the macro average F1 score is about 0.91.
Toponym Extraction and Disambiguation from Text: A Survey Windiastuti, Rizka; Krisnadhi, Adila Alfa; Budi, Indra
JOIV : International Journal on Informatics Visualization Vol 9, No 1 (2025)
Publisher : Society of Visual Informatics

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.62527/joiv.9.1.2763

Abstract

Toponym is an essential element of geospatial information. Traditionally, toponyms are collected in a gazetteer through field surveys that require significant resources, including labor, time, and money. Nowadays, we can utilize social media and online news portals to collect event locations or toponyms from the text. This article presents a survey of studies that focus on the extraction and disambiguation of toponyms from textual documents. While toponym extraction aims to identify toponyms from the text, toponym disambiguation determines their specific locations on the earth. The survey covered articles published between January 2015 and April 2023, presented in English, and gathered from five major journal databases. The survey was conducted by adopting the Kitchenham guidelines, consisting of an initial article search, article selection, and annotation process to facilitate the reporting phase. We employed Mendeley as a reference management tool and NVivo to categorize certain parts of the articles that are the focal points of interest in this survey. The primary focus of the survey was on the methods or approaches performed in the research articles to extract and disambiguate toponyms. Additionally, we also discuss some general challenges in toponym research, different applications for toponym extraction and disambiguation, data sources, and the use of languages other than English in the studies. The survey confirms that each approach has its limitations. Extracting and disambiguating toponyms from text is complex and challenging, especially for low-resource languages. We also suggest some research directions related to toponym extraction and disambiguation that could enrich the gazetteer.
A comparative study of natural language inference in Swahili using monolingual and multilingual models Faki Ali, Hajra; Alfa Krisnadhi, Adila
IAES International Journal of Artificial Intelligence (IJ-AI) Vol 14, No 2: April 2025
Publisher : Institute of Advanced Engineering and Science

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.11591/ijai.v14.i2.pp1597-1604

Abstract

Recent advancements in large language models (LLMs) have led to opportunities for improving applications across various domains. However, existing LLMs fine-tuned for Swahili or other African languages often rely on pre-trained multilingual models, resulting in a relatively small portion of training data dedicated to Swahili. In this study, we compare the performance of monolingual and multilingual models in Swahili natural language inference tasks using the cross-lingual natural language inference (XNLI) dataset. Our research demonstrates the superior effectiveness of dedicated Swahili monolingual models, achieving an accuracy rate of 69%. These monolingual models exhibit significantly enhanced precision, recall, and F1 scores, particularly in predicting contradiction and neutrality. Overall, the findings in this article emphasize the critical importance of using monolingual models in low-resource language processing contexts, providing valuable insights for developing more efficient and tailored natural language processing systems that benefit languages facing similar resource constraints.
Entity dan Relation Linking untuk Knowledge Graph Question Answering Menggunakan Pencarian Berjenjang Adila Alfa Krisnadhi; Mohammad Yani; Indra Budi
Jurnal Nasional Teknik Elektro dan Teknologi Informasi Vol 13 No 2: Mei 2024
Publisher : Departemen Teknik Elektro dan Teknologi Informasi, Fakultas Teknik, Universitas Gadjah Mada

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.22146/jnteti.v13i2.9184

Abstract

Knowledge graph question answering (KGQA) systems have an important role in retrieving data from a knowledge graph (KG). With the system, regular users can access data from a KG without the need to construct a formal SPARQL query. KGQA systems receive a natural language question (NLQ) and translate it into a SPARQL query through three main tasks, namely, entity and relation detection, entity and relation linking, and query construction. However, the translation is not trivial due to lexical gaps and entity ambiguity that may occur during entity or relation linking. This research proposed an approach based on multiclass classification of NLQ whose entity occurrences are detected into categories based on KG relations to address the lexical gap challenge. Next, to solve the entity ambiguity challenge, this research proposed a three-stage searching procedure to determine appropriate KG entities associated with the NLQ entities, given the correspondence between the NLQ and a particular KG relation. This three-stage searching consisted of text-based searching, vector-based searching, and entity and relation pairing. The proposed approach was evaluated on the SimpleQuestions and LC-QuAD 2.0 datasets. The experiments demonstrated that the proposed approach outperformed the state-of-the-art baseline. For the relation linking task, the proposed approach reached 89.87% and 74.83% recall for the SimpleQuestions and LC-QuAD 2.0, respectively. This approach also achieved 91.74% and 61.96% recall on the entity linking tasks for the SimpleQuestions and LC-QuAD 2.0, respectively.
Enhancing Table-to-Text Generation with Numerical Reasoning Using Graph2Seq Models Sulisetyo Puji Widodo; Adila Alfa Krisnadhi
International Journal of Innovation in Enterprise System Vol. 8 No. 2 (2024): International Journal of Innovation in Enterprise System
Publisher : School of Industrial and System Engineering, Telkom University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.25124/ijies.v8i02.236

Abstract

Interpreting data in tables into narratives is necessary because tables cannot explain their own data.Additionally, there is a need to produce more analytic narratives from the results of numericalreasoning on data from tables. The sequence-to-sequence (Seq2Seq) encoder-decoder structure is themost widely used in table-to-text generation (T2XG). However, Seq2Seq requires the linearization oftables, which can omit structural information and create hallucination problems. Alternatively, thegraph-to-sequence (Graph2Seq) encoder-decoder structure utilizes a graph encoder to better captureimportant data information. Several studies have shown that Graph2Seq outperforms Seq2Seq. Thus,this study applies Graph2Seq to T2XG, leveraging the structured nature of tables that can berepresented by graphs. This research initiates the use of Graph2Seq in T2XG with GCN-RNN andGraphSage-RNN, aiming to improve narrative generation from tables through enhanced numericalreasoning. Based on the automatic evaluation of the application of Graph2Seq on the T2XG task, ithas the same performance as the baseline model. Meanwhile, in human evaluation, Graphsage-RNNis better able to reduce the possibility of hallucinations in text compared to the baseline model andGCN-RNN.
Astronomical Image Denoising Using AttentionGAN Naufal, Faishal Zaka; Rachmadi, Muhammad Febrian; Krisnadhi, Adila Alfa
Syntax Literate Jurnal Ilmiah Indonesia
Publisher : Syntax Corporation

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.36418/syntax-literate.v10i1.13154

Abstract

Denoising astronomical images is a significant challenge in the field of astronomical data processing. Image data acquired from astronomical sources typically contains noise from various sources. The study aims to investigate the denoising of astronomical images using an image-to-image translation approach with AttentionGAN method. This method combines attention-guided techniques with a Generative Adversarial Network (GAN) model to improve the quality of noisy astronomical images. Attention-guided technique allows the model to learn the most important features of the image and guide the image generation process. This approach has been tested on several images in different domains, each with varying levels of noise. The results shows that AttentionGAN method produces denoised images with better and sharper quality than several other denoising methods. Two databases, The Panoramic Survey Telescope and Rapid Response System (PAN-STARRS) and the Sloan Digital Sky Survey (SDSS), were used in this research. Images acquired from PAN-STARRS contain noise, while images acquired from SDSS are clean. Overall, this research contributes to improving the quality of astronomical images by demonstrating the effectiveness of the AttentionGAN method in denoising noisy astronomical images. We employed denoising techniques using CycleGAN and AttentionGAN and evaluated them using metrics such as PSNR, SSIM, and FID. The analysis showed that the AttentionGAN model outperformed CycleGAN. We also conducted ablation studies to further investigate the components of the AttentionGAN model. This study provides a foundation for future research in the field of astronomical data processing, which has the potential to enhance image quality.
Model Klasifikasi Lightweight Untuk Deteksi Hama Pertanian Dengan Efficient NET, Spinal Net FC, dan Sharpness-Aware Minimization Pratama, Naufal Ihsan; Azizi, Fityan; Krisnadhi, Adila Alfa
Syntax Literate Jurnal Ilmiah Indonesia
Publisher : Syntax Corporation

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.36418/syntax-literate.v7i9.15119

Abstract

Petani Indonesia menghadapi tantangan besar yang disebabkan oleh hama, yang menyebabkan gagal panen, busuk batang, kerusakan daun, dan busuk buah. Mengembangkan model pendeteksian hama yang lightweight menjadi penting untuk membantu petani dalam peningkatan program pengendalian hama. Tujuan utama dari model ini adalah untuk mengklasifikasikan hama secara akurat dengan memanfaatkan kumpulan data berskala besar. Kumpulan data ini mencakup berbagai spesies dengan beragam skala, bentuk, latar belakang kompleks, dan tingkat kesamaan visual yang tinggi di antara spesies serangga. Penelitian ini menggunakan model klasifikasi lightweight berbasis Efficient Net. Model ini menggabungkan Spinal Net FC sebagai classifier dan mengadopsi Sharpness-Aware Minimization sebagai optimizer untuk meningkatkan kinerjanya. Model yang dikembangkan mencapai akurasi sebesar 68,2%. Selain itu, dengan mengimplementasikan metode yang diusulkan, performa model mengalami peningkatan yang signifikan, menghasilkan peningkatan akurasi tambahan sebesar 4%.