{"id":4975,"date":"2023-07-17T11:24:15","date_gmt":"2023-07-17T03:24:15","guid":{"rendered":"https:\/\/inventec2.mjitec.tw\/?page_id=4975"},"modified":"2025-02-03T09:17:28","modified_gmt":"2025-02-03T01:17:28","slug":"ai","status":"publish","type":"page","link":"https:\/\/inventec2.mjitec.tw\/en\/ai\/","title":{"rendered":"AI Technology"},"content":{"rendered":"

[vc_row full_width=”stretch_row” css=”.vc_custom_1657857836164{border-bottom-width: 0px !important;padding-top: 0px !important;background-color: #efefef !important;}” el_id=”what-we-do”][vc_column css=”.vc_custom_1657792365832{padding-top: 0px !important;padding-bottom: 0px !important;}” el_class=”m_p”]

\r\n
<\/div>\t\t\t\r\n\t\t\t<\/div>\n
\n \t
\n \t\t\n\t \n\t

Applications <\/h2>\n\t <\/div><\/div>
\r\n
<\/div>\t\t\t\r\n\t\t\t<\/div>[vc_tta_tabs section_title_tag=”h3″ color=”white” alignment=”center” active_section=”1″ el_class=”custom-tab”][vc_tta_section title=”All” tab_id=”all”][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2521″ img_size=”full” alignment=”center” css=”.vc_custom_1656582527742{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542922353{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

A Mixed-Domain Self-Attention Network for Multilabel Cardiac Irregularity Classification Using Reduced-Lead Electrocardiogram<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657000397990{padding-right: 40px !important;padding-left: 40px !important;}”]As part of the PhysioNet Computing in Cardiology Challenge 2021, our team HaoWan AIeC, proposed Mixed-Domain Self-Attention Resnet (MDARsn) to identify cardiac abnormalities from reduced-lead ECG.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2522″ img_size=”full” alignment=”center” css=”.vc_custom_1657000811717{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542941656{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Transition Motion Tensor: A Data-Driven Approach for Versatile and Controllable Agents in Physically Simulated Environments<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657000855309{padding-right: 40px !important;padding-left: 40px !important;}”]This paper proposes the Transition Motion Tensor, a data-driven framework that creates novel and physically accurate transitions outside of the motion dataset.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2519″ img_size=”full” alignment=”center” css=”.vc_custom_1657000885041{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542961173{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

CARL: Controllable Agent with Reinforcement Learning for Quadruped Locomotion<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657002086228{padding-right: 40px !important;padding-left: 40px !important;}”]In this paper, we present CARL, a quadruped agent that can be controlled with high-level directives and react naturally to dynamic environments.<\/span>[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2516″ img_size=”full” alignment=”center” css=”.vc_custom_1657000938185{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542969306{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

A Trainable Reconciliation Method for Hierarchical Time-Series<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657000964075{padding-right: 40px !important;padding-left: 40px !important;}”]In this paper, we propose a new general, flexible, and easy-to-implement reconciliation strategy based on an encoder-decoder neural network.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2515″ img_size=”full” alignment=”center” css=”.vc_custom_1657000982393{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542979872{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Demystifying data and AI for manufacturing: case studies from a major computer maker<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001008526{padding-right: 40px !important;padding-left: 40px !important;}”]We create a deep learning-based algorithm for visual inspection of product appearances, which requires significantly less defect training data compared to traditional approaches.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2514″ img_size=”full” alignment=”center” css=”.vc_custom_1657001028338{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542993944{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

TrustMAE: A Noise-Resilient Defect Classification Framework using Memory-Augmented Auto-Encoders with Trust Regions<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001053828{padding-right: 40px !important;padding-left: 40px !important;}”]In this paper, we propose a framework called Trust MAE to address the problem of product defect classification. Instead of relying on defective images that are difficult to collect and laborious to label, our framework can accept datasets with unlabeled images.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2517″ img_size=”full” alignment=”center” css=”.vc_custom_1657001068948{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660543002810{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Domain-Generalized Textured Surface Anomaly Detection<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001086927{padding-right: 40px !important;padding-left: 40px !important;}”]By observing normal and abnormal surface data across multiple source domains, our model is expected to be generalized to an unseen textured surface of interest, in which only a small number of normal data can be observed during testing.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2520″ img_size=”full” alignment=”center” css=”.vc_custom_1657001117669{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660543010524{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

A Dense Tensor Accelerator with Data Exchange Mesh for DNN and Vision Workloads<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001144036{padding-right: 40px !important;padding-left: 40px !important;}”]We propose a dense tensor accelerator called VectorMesh, a scalable, memory-efficient architecture that can support a wide variety of DNN and computer vision workloads.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2518″ img_size=”full” alignment=”center” css=”.vc_custom_1657001164169{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660543018556{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

MERIT: Tensor transform for memory-efficient vision processing on parallel architectures<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001897002{padding-right: 40px !important;padding-left: 40px !important;}”]We propose a mathematical formulation which can be useful for transferring the parallel algorithm optimization knowledge across computing platforms.<\/span>[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2513″ img_size=”full” alignment=”center” css=”.vc_custom_1657001206250{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660543030728{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

GrateTile: Efficient Sparse Tensor Tiling for CNN Processing<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001228899{padding-right: 40px !important;padding-left: 40px !important;}”]We propose GrateTile, an efficient, hardware friendly data storage scheme for sparse CNN feature maps (activations).[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5039″ img_size=”full” alignment=”center” css=”.vc_custom_1689651075408{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689571400546{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Configuration through Optimization for In-Memory Computing Hardware and Simulators<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648625896{padding-right: 40px !important;padding-left: 40px !important;}”]In this paper, we aim to simulate unknown IMC hardware with existing simulators to reduce the development time for a new simulator. [\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5042″ img_size=”full” alignment=”center” css=”.vc_custom_1689651083523{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689574357289{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

FedTrust: Towards Building
\nSecure Robust and Trustworthy Moderators for Federated Learning<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648657562{padding-right: 40px !important;padding-left: 40px !important;}”]Most Federated Learning (FL) systems are built upon a strong assumption of trust—clients fully trust the centralized moderator, which might not be feasible in practice. This work aims to mitigate the assumption by using appropriate cryptographic tools.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”13109″ img_size=”full” alignment=”center” css=”.vc_custom_1695890269305{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689574470490{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

A Robust Collaborative Learning Framework Using Data Digests and Synonyms to Represent Absent Clients<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648681478{padding-right: 40px !important;padding-left: 40px !important;}”]We propose Collaborative Learning with Synonyms (CLSyn), a robust and versatile collaborative machine learning framework that can tolerate unexpected client absence during training while maintaining high model accuracy. [\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5018″ img_size=”full” alignment=”center” css=”.vc_custom_1689651105551{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689574555124{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

USAID: Intelligent Forecasting Competition<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648702930{padding-right: 40px !important;padding-left: 40px !important;}”]The competition focused on the global health topic, where we need to build an AI algorithm to provide an accurate prediction for contraceptive consumption in Africa, specifically Cote d’Ivoire.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5021″ img_size=”full” alignment=”center” css=”.vc_custom_1689651114053{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689574641580{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Expanding Versatility of Agile Locomotion through Policy Transitions Using Latent State Representation<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648725862{padding-right: 40px !important;padding-left: 40px !important;}”]This paper proposes the transition-net, a robust transition strategy that expands the versatility of robot locomotion in the real-world setting.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5024″ img_size=”full” alignment=”center” css=”.vc_custom_1689651122877{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689574723013{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Interpretable Estimation of the risk of Heart Failure Hospitalization from a 30-second Electrocardiogram<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648778401{padding-right: 40px !important;padding-left: 40px !important;}”]This study shows it is possible to estimate hospitalization for congestive heart failure by a 30 seconds single-lead electrocardiogram signal. Using a machine learning approach not only results in greater predictive power but also provides clinically meaningful interpretations. [\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5027″ img_size=”full” alignment=”center” css=”.vc_custom_1689651132827{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689574792350{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

A Benchmark for Machine-Learning based Non-Invasive Blood Pressure Estimation using Photoplethysmogram<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648813381{padding-right: 40px !important;padding-left: 40px !important;}”]Blood Pressure (BP) is an important cardiovascular health indicator. BP is usually monitored non-invasively with a cuff-based device, which can be bulky and inconvenient. Thus, continuous and portable BP monitoring devices, such as those based on a photoplethysmography (PPG) waveform, are desirable. In particular, Machine Learning (ML) based BP estimation approaches have gained considerable attention as they have the potential to estimate intermittent or continuous BP with only a single PPG measurement. [\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”67219″ img_size=”full” alignment=”center” css=”.vc_custom_1705470572971{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1705470592489{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Improving Limited Supervised Foot Ulcer Segmentation Using Cross-Domain Augmentation<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1705470679683{padding-right: 40px !important;padding-left: 40px !important;}”]Diabetic foot ulcers pose health risks. Our method, TransMix, enhances wound segmentation data, improving model training with limited annotated images, leading to significant performance improvement.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5033″ img_size=”full” alignment=”center” css=”.vc_custom_1689651185845{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689574925233{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Development of a Deep Learning-Based Tool to Assist Wound Classification<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648855867{padding-right: 40px !important;padding-left: 40px !important;}”]This paper presents a deep learning-based wound classification tool that can assist medical personnel in non-wound care specialization to classify five key wound conditions, namely deep wound, infected wound, arterial wound, venous wound, and pressure wound, given color images captured using readily available cameras. The accuracy of the classification is vital for appropriate wound management.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5036″ img_size=”full” alignment=”center” css=”.vc_custom_1689651194408{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689575107021{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

BARN Challenge 2023 – Navigation in Tighly Constrained Spaces<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648879168{padding-right: 40px !important;padding-left: 40px !important;}”]Many existing navigation systems allow robots to move from one point to another in a collision-free manner, which may create the impression that navigation is a solved problem. However, autonomous mobile robots still struggle in many scenarios, e.g., colliding with or getting stuck in novel and tightly constrained spaces. [\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”67153″ img_size=”full” alignment=”center” css=”.vc_custom_1705472065766{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1705472008507{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

An Efficient CKKS-FHEW\/TFHE Hybrid Encrypted Inference Framework<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1705472277252{padding-right: 40px !important;padding-left: 40px !important;}”]We aim to investigate the opportunity to apply the CKKS-FHEW\/TFHE hybrid approach to NNs.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1709884336009{background-color: #ffffff !important;}”][vc_single_image image=”67555″ img_size=”full” alignment=”center” css=”.vc_custom_1709884326815{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1709884348941{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Multi-modal Heart Failure Risk Estimation based on Short ECG and Sampled Long-Term HRV<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1709884362778{padding-right: 40px !important;padding-left: 40px !important;}”]This paper propose several multi-modal approaches that combine 30-s ECG recordings and approximate long-term Heart Rate Variability (HRV) data to estimate the risk of HF hospitalization. The experiments demonstrate high model performance for HF risk assessment. Our approach implemented in the myHeartScore App offers cost-effective and highly accessible HF risk assessment, contributing to its prevention and management.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”68461″ img_size=”full” alignment=”center” css=”.vc_custom_1713840064248{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1713840085552{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Expert Composer Policy: Scalable Skill Repertoire for Quadruped Robots<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1713840102360{padding-right: 40px !important;padding-left: 40px !important;}”]A framework to reliably expand the skill repertoire of quadruped agents. The composer policy links pair of experts via transitions to a sampled target state, allowing experts to be composed sequentially.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1721033339406{background-color: #ffffff !important;}”][vc_single_image image=”68599″ img_size=”full” alignment=”center” css=”.vc_custom_1721033259157{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1721033273102{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Learning with Instance-dependent Noisy Labels by Anchor Hallucination and Hard Sample Label Correction<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1721033287503{padding-right: 40px !important;padding-left: 40px !important;}”]A framework aims to solve the real-world noisy label issues by explicitly distinguishing between clean vs. noisy and easy vs. hard samples.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”69123″ img_size=”full” alignment=”center” css=”.vc_custom_1727088442911{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1727088451860{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Learning Multi-Manifold Embedding for Out-Of-Distribution Detection<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1727088470862{padding-right: 40px !important;padding-left: 40px !important;}”]Our Multi-Manifold Embedding Learning framework improves out-of-distribution (OOD) detection by jointly optimizing hypersphere and hyperbolic spaces. We also investigate a test-time enrollment strategy for real-world usages.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1721033339406{background-color: #ffffff !important;}”][vc_single_image image=”69141″ img_size=”full” alignment=”center” css=”.vc_custom_1727088512006{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1727088519963{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Dual Deep Learning System to Digitize and Classify 12-lead ECGs from Scanned Images<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1727088533158{padding-right: 40px !important;padding-left: 40px !important;}”]As part of the PhysioNet\/Computing in Cardiology Challenge 2024, our team, Inventec AIC, developed a dual deep-learning system to digitize and classify 12-lead electrocardiograms (ECG) from scanned images. Our approach was awarded 2nd place in the classification task of the PhysioNet Challenge.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”69154″ img_size=”full” alignment=”center” css=”.vc_custom_1727088573934{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1727088580521{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Profiling Patient Transcript Using Large Language Model Reasoning Augmentation for Alzheimer\u2019s Disease Detection<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1727088592255{padding-right: 40px !important;padding-left: 40px !important;}”]This idea uses LLM to generate patient’s linguistic profile, which is helpful in assessing Alzheimer\u2019s Disease. Method: designing prompt to instruct LLM to identify language deficits from participant’s transcript Evaluation: 8.34% performance gain (measured in accuracy).[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1729846196939{background-color: #ffffff !important;}”][vc_single_image image=”69214″ img_size=”full” alignment=”center” css=”.vc_custom_1729846248821{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1729846263999{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Benchmarking Smoothness and Reducing High-Frequency Oscillations in Continuous Control Policies<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1729846277594{padding-right: 40px !important;padding-left: 40px !important;}”]In this paper, we analyze methods for reducing high-frequency oscillations in deep RL, focusing on loss regularization and architectural approaches. We introduce hybrid methods combining both techniques, with our best hybrid improving control smoothness by 26.8% while only degrading performance by 2.8%.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”69967″ img_size=”full” alignment=”center” css=”.vc_custom_1737714371067{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1737714316346{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Who Brings the Frisbee: Probing Hidden Hallucination Factors in Large Vision-Language Model via Causality Analysis<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1737714335060{padding-right: 40px !important;padding-left: 40px !important;}”]An causal analytical framework identifies and addresses hallucination in vision language models.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″][\/vc_column_inner][\/vc_row_inner][\/vc_tta_section][vc_tta_section title=”Smart Health” tab_id=”smart_health”][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2521″ img_size=”full” alignment=”center” css=”.vc_custom_1656582527742{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542922353{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

A Mixed-Domain Self-Attention Network for Multilabel Cardiac Irregularity Classification Using Reduced-Lead Electrocardiogram<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657000397990{padding-right: 40px !important;padding-left: 40px !important;}”]As part of the PhysioNet Computing in Cardiology Challenge 2021, our team HaoWan AIeC, proposed Mixed-Domain Self-Attention Resnet (MDARsn) to identify cardiac abnormalities from reduced-lead ECG.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689658777104{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5024″ img_size=”full” alignment=”center” css=”.vc_custom_1689651122877{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689574723013{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Interpretable Estimation of the risk of Heart Failure Hospitalization from a 30-second Electrocardiogram<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648778401{padding-right: 40px !important;padding-left: 40px !important;}”]This study shows it is possible to estimate hospitalization for congestive heart failure by a 30 seconds single-lead electrocardiogram signal. Using a machine learning approach not only results in greater predictive power but also provides clinically meaningful interpretations. [\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5027″ img_size=”full” alignment=”center” css=”.vc_custom_1689651132827{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689574792350{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

A Benchmark for Machine-Learning based Non-Invasive Blood Pressure Estimation using Photoplethysmogram<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648813381{padding-right: 40px !important;padding-left: 40px !important;}”]Blood Pressure (BP) is an important cardiovascular health indicator. BP is usually monitored non-invasively with a cuff-based device, which can be bulky and inconvenient. Thus, continuous and portable BP monitoring devices, such as those based on a photoplethysmography (PPG) waveform, are desirable. In particular, Machine Learning (ML) based BP estimation approaches have gained considerable attention as they have the potential to estimate intermittent or continuous BP with only a single PPG measurement. [\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689658777104{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5033″ img_size=”full” alignment=”center” css=”.vc_custom_1689651185845{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689574925233{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Development of a Deep Learning-Based Tool to Assist Wound Classification<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648855867{padding-right: 40px !important;padding-left: 40px !important;}”]This paper presents a deep learning-based wound classification tool that can assist medical personnel in non-wound care specialization to classify five key wound conditions, namely deep wound, infected wound, arterial wound, venous wound, and pressure wound, given color images captured using readily available cameras. The accuracy of the classification is vital for appropriate wound management.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”67219″ img_size=”full” alignment=”center” css=”.vc_custom_1705470775705{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1705470762834{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Improving Limited Supervised Foot Ulcer Segmentation Using Cross-Domain Augmentation<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1705470747682{padding-right: 40px !important;padding-left: 40px !important;}”]Diabetic foot ulcers pose health risks. Our method, TransMix, enhances wound segmentation data, improving model training with limited annotated images, leading to significant performance improvement.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1709884454681{background-color: #ffffff !important;}”][vc_single_image image=”67555″ img_size=”full” alignment=”center” css=”.vc_custom_1709884445047{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1711436590672{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Multi-modal Heart Failure Risk Estimation based on Short ECG and Sampled Long-Term HRV<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1709884472952{padding-right: 40px !important;padding-left: 40px !important;}”]This paper propose several multi-modal approaches that combine 30-s ECG recordings and approximate long-term Heart Rate Variability (HRV) data to estimate the risk of HF hospitalization. The experiments demonstrate high model performance for HF risk assessment. Our approach implemented in the myHeartScore App offers cost-effective and highly accessible HF risk assessment, contributing to its prevention and management.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”69141″ img_size=”full” alignment=”center” css=”.vc_custom_1727088826851{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1727088782734{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Dual Deep Learning System to Digitize and Classify 12-lead ECGs from Scanned Images<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1727088850885{padding-right: 40px !important;padding-left: 40px !important;}”]As part of the PhysioNet\/Computing in Cardiology Challenge 2024, our team, Inventec AIC, developed a dual deep-learning system to digitize and classify 12-lead electrocardiograms (ECG) from scanned images. Our approach was awarded 2nd place in the classification task of the PhysioNet Challenge.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1709884454681{background-color: #ffffff !important;}”][vc_single_image image=”69154″ img_size=”full” alignment=”center” css=”.vc_custom_1727088792085{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1727088746237{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Profiling Patient Transcript Using Large Language Model Reasoning Augmentation for Alzheimer\u2019s Disease Detection<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1727088732367{padding-right: 40px !important;padding-left: 40px !important;}”]This idea uses LLM to generate patient’s linguistic profile, which is helpful in assessing Alzheimer\u2019s Disease. Method: designing prompt to instruct LLM to identify language deficits from participant’s transcript Evaluation: 8.34% performance gain (measured in accuracy).[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5020″ img_size=”full” alignment=”center” css=”.vc_custom_1721291652249{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1721291671465{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

USAID: Intelligent Forecasting Competition<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1721291686709{padding-right: 40px !important;padding-left: 40px !important;}”]The competition focused on the global health topic, where we need to build an AI algorithm to provide an accurate prediction for contraceptive consumption in Africa, specifically Cote d’Ivoire.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][\/vc_tta_section][vc_tta_section title=”Smart Manufacturing” tab_id=”manufacturing”][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2516″ img_size=”full” alignment=”center” css=”.vc_custom_1657000938185{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542969306{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

A Trainable Reconciliation Method for Hierarchical Time-Series<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657000964075{padding-right: 40px !important;padding-left: 40px !important;}”]In this paper, we propose a new general, flexible, and easy-to-implement reconciliation strategy based on an encoder-decoder neural network.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2517″ img_size=”full” alignment=”center” css=”.vc_custom_1657001068948{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660543002810{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Domain-Generalized Textured Surface Anomaly Detection<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001086927{padding-right: 40px !important;padding-left: 40px !important;}”]By observing normal and abnormal surface data across multiple source domains, our model is expected to be generalized to an unseen textured surface of interest, in which only a small number of normal data can be observed during testing.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2515″ img_size=”full” alignment=”center” css=”.vc_custom_1657000982393{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542979872{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Demystifying data and AI for manufacturing: case studies from a major computer maker<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001008526{padding-right: 40px !important;padding-left: 40px !important;}”]We create a deep learning-based algorithm for visual inspection of product appearances, which requires significantly less defect training data compared to traditional approaches.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2514″ img_size=”full” alignment=”center” css=”.vc_custom_1657001028338{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542993944{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

TrustMAE: A Noise-Resilient Defect Classification Framework using Memory-Augmented Auto-Encoders with Trust Regions<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001053828{padding-right: 40px !important;padding-left: 40px !important;}”]In this paper, we propose a framework called Trust MAE to address the problem of product defect classification. Instead of relying on defective images that are difficult to collect and laborious to label, our framework can accept datasets with unlabeled images.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”69970″ img_size=”full” alignment=”center” css=”.vc_custom_1737714515530{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1737714472404{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Learning Multi-Manifold Embedding for Out-Of-Distribution Detection<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1737714501728{padding-right: 40px !important;padding-left: 40px !important;}”]Our Multi-Manifold Embedding Learning framework improves out-of-distribution (OOD) detection by jointly optimizing hypersphere and hyperbolic spaces. We also investigate a test-time enrollment strategy for real-world usages.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1737714456998{margin-bottom: 30px !important;padding-bottom: 20px !important;}”][\/vc_column_inner][\/vc_row_inner][\/vc_tta_section][vc_tta_section title=”Robotics” tab_id=”robotics”][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2522″ img_size=”full” alignment=”center” css=”.vc_custom_1657000811717{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542941656{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Transition Motion Tensor: A Data-Driven Approach for Versatile and Controllable Agents in Physically Simulated Environments<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657000855309{padding-right: 40px !important;padding-left: 40px !important;}”]This paper proposes the Transition Motion Tensor, a data-driven framework that creates novel and physically accurate transitions outside of the motion dataset.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689658606633{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2519″ img_size=”full” alignment=”center” css=”.vc_custom_1657000885041{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542961173{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

CARL: Controllable Agent with Reinforcement Learning for Quadruped Locomotion<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657002086228{padding-right: 40px !important;padding-left: 40px !important;}”]In this paper, we present CARL, a quadruped agent that can be controlled with high-level directives and react naturally to dynamic environments.<\/span>[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1713840181513{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5021″ img_size=”full” alignment=”center” css=”.vc_custom_1689651114053{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1689574641580{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Expanding Versatility of Agile Locomotion through Policy Transitions Using Latent State Representation<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1689648725862{padding-right: 40px !important;padding-left: 40px !important;}”]This paper proposes the transition-net, a robust transition strategy that expands the versatility of robot locomotion in the real-world setting.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1713840188758{background-color: #ffffff !important;}”][vc_single_image image=”68461″ img_size=”full” alignment=”center” css=”.vc_custom_1713840222430{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1713840239451{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Expert Composer Policy: Scalable Skill Repertoire for Quadruped Robots<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1713840250034{padding-right: 40px !important;padding-left: 40px !important;}”]A framework to reliably expand the skill repertoire of quadruped agents. The composer policy links pair of experts via transitions to a sampled target state, allowing experts to be composed sequentially.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1713840181513{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5038″ img_size=”full” alignment=”center” css=”.vc_custom_1721291791324{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1721291797917{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

BARN Challenge 2023 – Navigation in Tighly Constrained Spaces<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1721291812260{padding-right: 40px !important;padding-left: 40px !important;}”]Many existing navigation systems allow robots to move from one point to another in a collision-free manner, which may create the impression that navigation is a solved problem. However, autonomous mobile robots still struggle in many scenarios, e.g., colliding with or getting stuck in novel and tightly constrained spaces.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1729846324532{background-color: #ffffff !important;}”][vc_single_image image=”69214″ img_size=”full” alignment=”center” css=”.vc_custom_1729846410189{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1729846482971{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Benchmarking Smoothness and Reducing High-Frequency Oscillations in Continuous Control Policies<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1729846399112{padding-right: 40px !important;padding-left: 40px !important;}”]In this paper, we analyze methods for reducing high-frequency oscillations in deep RL, focusing on loss regularization and architectural approaches. We introduce hybrid methods combining both techniques, with our best hybrid improving control smoothness by 26.8% while only degrading performance by 2.8%.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][\/vc_tta_section][vc_tta_section title=”NPU IP” tab_id=”NPU_IP”][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2520″ img_size=”full” alignment=”center” css=”.vc_custom_1657001117669{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660543010524{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

A Dense Tensor Accelerator with Data Exchange Mesh for DNN and Vision Workloads<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001144036{padding-right: 40px !important;padding-left: 40px !important;}”]We propose a dense tensor accelerator called VectorMesh, a scalable, memory-efficient architecture that can support a wide variety of DNN and computer vision workloads.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564498373{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2513″ img_size=”full” alignment=”center” css=”.vc_custom_1657001206250{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660543030728{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

GrateTile: Efficient Sparse Tensor Tiling for CNN Processing<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001228899{padding-right: 40px !important;padding-left: 40px !important;}”]We propose GrateTile, an efficient, hardware friendly data storage scheme for sparse CNN feature maps (activations).[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2518″ img_size=”full” alignment=”center” css=”.vc_custom_1657001164169{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660543018556{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

MERIT: Tensor transform for memory-efficient vision processing on parallel architectures<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1657001897002{padding-right: 40px !important;padding-left: 40px !important;}”]We propose a mathematical formulation which can be useful for transferring the parallel algorithm optimization knowledge across computing platforms.<\/span>[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][\/vc_tta_section][vc_tta_section title=”Trustworthy AI” tab_id=”trustworthy_ai”][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”68599″ img_size=”full” alignment=”center” css=”.vc_custom_1721032915421{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1721032947877{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Learning with Instance-dependent Noisy Labels by Anchor Hallucination and Hard Sample Label Correction<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1721032961034{padding-right: 40px !important;padding-left: 40px !important;}”]A framework aims to solve the real-world noisy label issues by explicitly distinguishing between clean vs. noisy and easy vs. hard samples.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1721291168836{background-color: #ffffff !important;}”][vc_single_image image=”5041″ img_size=”full” alignment=”center” css=”.vc_custom_1721290935354{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1721290952791{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Configuration through Optimization for In-Memory Computing Hardware and Simulators<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1721290969718{padding-right: 40px !important;padding-left: 40px !important;}”]In this paper, we aim to simulate unknown IMC hardware with existing simulators to reduce the development time for a new simulator.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1721291174582{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”5008″ img_size=”full” alignment=”center” css=”.vc_custom_1721291012058{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1721291044005{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

FedTrust: Towards Building
\nSecure Robust and Trustworthy Moderators for Federated Learning<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1721291029241{padding-right: 40px !important;padding-left: 40px !important;}”]Most Federated Learning (FL) systems are built upon a strong assumption of trust—clients fully trust the centralized moderator, which might not be feasible in practice. This work aims to mitigate the assumption by using appropriate cryptographic tools.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1721291181329{background-color: #ffffff !important;}”][vc_single_image image=”13109″ img_size=”full” alignment=”center” css=”.vc_custom_1721291457443{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1721291090774{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

A Robust Collaborative Learning Framework Using Data Digests and Synonyms to Represent Absent Clients<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1721291128396{padding-right: 40px !important;padding-left: 40px !important;}”]We propose Collaborative Learning with Synonyms (CLSyn), a robust and versatile collaborative machine learning framework that can tolerate unexpected client absence during training while maintaining high model accuracy.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1721291174582{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”67153″ img_size=”full” alignment=”center” css=”.vc_custom_1721291527945{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1721291543120{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

An Efficient CKKS-FHEW\/TFHE Hybrid Encrypted Inference Framework\u200b<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1721291557597{padding-right: 40px !important;padding-left: 40px !important;}”]We aim to investigate the opportunity to apply the CKKS-FHEW\/TFHE hybrid approach to NNs.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1721291181329{background-color: #ffffff !important;}”][vc_single_image image=”69967″ img_size=”full” alignment=”center” css=”.vc_custom_1737714668282{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1737714639860{padding-right: 40px !important;padding-left: 40px !important;}”]<\/p>\n

Who Brings the Frisbee: Probing Hidden Hallucination Factors in Large Vision-Language Model via Causality Analysis<\/h3>\n

[\/vc_column_text][vc_column_text css=”.vc_custom_1737714653596{padding-right: 40px !important;padding-left: 40px !important;}”]An causal analytical framework identifies and addresses hallucination in vision language models.[\/vc_column_text]\r\n\r\n\r\n

\r\n\tLearn more<\/a>\r\n<\/div>\r\n\r\n[\/vc_column_inner][\/vc_row_inner][\/vc_tta_section][\/vc_tta_tabs][\/vc_column][\/vc_row]<\/p>\n<\/div>","protected":false},"excerpt":{"rendered":"

[vc_row full_width=”stretch_row” css=”.vc_custom_1657857836164{border-bottom-width: 0px !important;padding-top: 0px !important;background-color: #efefef !important;}” el_id=”what-we-do”][vc_column css=”.vc_custom_1657792365832{padding-top: 0px !important;padding-bottom: 0px !important;}” el_class=”m_p”][vc_tta_tabs section_title_tag=”h3″ color=”white” alignment=”center” active_section=”1″ el_class=”custom-tab”][vc_tta_section title=”All” tab_id=”all”][vc_row_inner el_class=”md-two-col wrap_flex”][vc_column_inner el_class=”card_box” width=”1\/2″ css=”.vc_custom_1689564470008{margin-bottom: 30px !important;padding-bottom: 20px !important;background-color: #ffffff !important;}”][vc_single_image image=”2521″ img_size=”full” alignment=”center” css=”.vc_custom_1656582527742{background-color: #ffffff !important;}”][vc_column_text css=”.vc_custom_1660542922353{padding-right: 40px !important;padding-left: 40px !important;}”] A Mixed-Domain Self-Attention Network for Multilabel Cardiac Irregularity Classification Using…<\/p>\n","protected":false},"author":1,"featured_media":0,"parent":0,"menu_order":1,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-4975","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/inventec2.mjitec.tw\/en\/wp-json\/wp\/v2\/pages\/4975","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/inventec2.mjitec.tw\/en\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/inventec2.mjitec.tw\/en\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/inventec2.mjitec.tw\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/inventec2.mjitec.tw\/en\/wp-json\/wp\/v2\/comments?post=4975"}],"version-history":[{"count":18,"href":"https:\/\/inventec2.mjitec.tw\/en\/wp-json\/wp\/v2\/pages\/4975\/revisions"}],"predecessor-version":[{"id":69998,"href":"https:\/\/inventec2.mjitec.tw\/en\/wp-json\/wp\/v2\/pages\/4975\/revisions\/69998"}],"wp:attachment":[{"href":"https:\/\/inventec2.mjitec.tw\/en\/wp-json\/wp\/v2\/media?parent=4975"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}