公开(公告)号：WO2021245072A1

公开(公告)日：2021-12-09

申请号：PCT/EP2021/064662

申请日：2021-06-01

Applicant: IMEC VZW ,  VRIJE UNIVERSITEIT BRUSSEL

Inventor： ABRAHAMYAN, Lusine ,  DELIGIANNIS, Nikolaos

IPC: G06N3/04 ,  G06N3/08 ,  G06N3/063 ,  G06N3/0445 ,  G06N3/0454 ,  G06N3/084 ,  G06N3/088

Abstract: The present disclosure relates to a computer implemented method for training a learning model by means of a distributed learning system comprising computing nodes, the computing nodes respectively implementing the learning model and deriving a gradient information for updating the learning model based on training data, the method comprising: encoding, by the computing nodes, the gradient information by exploiting a correlation across the gradient information from the respective computing nodes; exchanging, by the computing nodes, the encoded gradient information within the distributed learning system; determining an aggregate gradient information based on the encoded gradient information from the computing nodes; and updating the learning model of the computing nodes with the aggregate gradient information, thereby training the learning model.

公开(公告)号：WO2022167304A1

公开(公告)日：2022-08-11

申请号：PCT/EP2022/051835

申请日：2022-01-27

Applicant: IMEC VZW ,  VRIJE UNIVERSITEIT BRUSSEL

Inventor： DELIGIANNIS, Nikolaos ,  JOUKOVSKY, Boris

IPC: G06F16/35 ,  G06F40/00 ,  G06F40/30 ,  G06N3/02

Abstract: A disinformation detection system (100) for verification of a plurality of information items comprises:- a scoring microservice (103; 300; 800) configured to execute at least one trained machine learning model (301, 302; 801, 802) adapted to generate a disinformation prediction for each information item; and- a graph neural network (303, 803), abbreviated GNN, configured to iteratively update the disinformation prediction through iterations of a mean field algorithm exploiting correlations between the information items as modelled in a Markov Random Field (807), abbreviated MRF. In the MRF the information items represent nodes, disinformation predictions represent node labels, and correlations represent edge values.The disinformation detection system (100) further comprises a relevancy unit (820, 821) configured to determine a relevancy value (R) for a path between a starting information item and a neighbouring information item in the MRF (807), and to output the relevancy value (R) as an influence contribution of the neighbouring information item to the disinformation prediction of the starting information item.

公开(公告)号：WO2022167302A1

公开(公告)日：2022-08-11

申请号：PCT/EP2022/051831

申请日：2022-01-27

Applicant: IMEC VZW ,  VRIJE UNIVERSITEIT BRUSSEL

Inventor： DELIGIANNIS, Nikolaos ,  HUU DO, Tien ,  BERNEMAN, Marc ,  VANDEN BROUCKE, Steven ,  CERNEJ, Vitalijus

IPC: G06F16/35 ,  G06F40/30 ,  G06Q30/00 ,  G06Q30/02 ,  G06Q30/06 ,  G06Q50/00

Abstract: A disinformation detection system (100) for automated verification of information items comprises an event bus (110) and event-driven microservices (101-106). The event-driven microservices (101-106) comprise: - a scoring microservice (103; 300) configured to execute at least one trained machine learning model (301, 302, 303, 305) adapted to generate a disinformation prediction for each information item; - a training microservice (104) configured to train the at least one machine learning model (301, 302, 303, 305) based on input obtained from researchers; and - a monitoring microservice (105; 200) configured to obtain the information items and related data, and forward them to the scoring microservice (103; 300). The monitoring microservice (105; 200) comprises: - a data storage (240); - at least one background harvester (235, 236, 237) configured to periodically fetch and store information items and/or related data from a particular information source; and - at least one on-demand harvester (221, 222) configured to fetch and store an information item and/or related data in return to a URL or other type of query of the information item.

公开(公告)号：WO2012072637A1

公开(公告)日：2012-06-07

申请号：PCT/EP2011/071296

申请日：2011-11-29

Applicant: IBBT ,  VRIJE UNIVERSITEIT BRUSSEL ,  DELIGIANNIS, Nikolaos ,  MUNTEANU, Adrian ,  BARBARIEN, Joeri

Inventor： DELIGIANNIS, Nikolaos ,  MUNTEANU, Adrian ,  BARBARIEN, Joeri

IPC: H04N7/26

CPC classification number: H04N19/00533 ,  H04N19/115 ,  H04N19/13 ,  H04N19/152 ,  H04N19/159 ,  H04N19/172 ,  H04N19/30 ,  H04N19/395 ,  H04N19/44 ,  H04N19/46

Abstract: The present invention is related to a hash-based distributed video coding architecture. At the encoder, the input video sequence is organized in Groups of Pictures (GOPs) and is decomposed into key frames, i.e., the first frame in each GOP, and WZ frames. The key frames are encoded using H264/AVC Intra frame coding. The Wyner-Ziv (WZ) frames are encoded in two parts, a hash layer and a WZ layer. To construct the hash information, the WZ frames are quantized and each quantized frame is then decorrelated using spatio-temporal prediction and entropy coded, and then multiplexed with the encoded key frames. At the decoder, the intra bit stream is H264/AVC decoded and the intra frames are stored in a reference frame buffer. The hash is decoded by inverting the tasks applied at the encoder, i.e. entropy decoding and inverse spatio-temporal prediction, and the obtained bit planes are stored. Next, Overlapped Block Motion Estimation and Probalistic Compensation (OBMEPC) is used to estimate the missing bit planes in the side - information. The decoder utilizes the hash information and the side information frame created by OBMEPC to perform online estimation of the correlation channel, and produces soft estimates used to decode the WZ bit planes.

Abstract translation: 本发明涉及一种基于散列的分布式视频编码架构。 在编码器处，输入视频序列被组织成图像组（GOP），并被分解成关键帧，即每个GOP中的第一帧和WZ帧。 关键帧使用H264 / AVC帧内编码进行编码。 Wyner-Ziv（WZ）帧被分成两部分，一个散列层和一个WZ层。 为了构建散列信息，对WZ帧进行量化，然后使用时空预测和熵编码对每个量化帧进行去相关，然后与编码的关键帧进行复用。 在解码器中，内部比特流被解码为H264 / AVC，帧内存储在参考帧缓冲器中。 通过反转在编码器处应用的任务，即熵解码和反时空预测来解码散列，并且存储所获得的位平面。 接下来，重叠块运动估计和平衡补偿（OBMEPC）用于估计侧信息中的丢失位平面。 解码器利用由OBMEPC创建的散列信息和边信息帧来执行相关信道的在线估计，并产生用于解码WZ位平面的软估计。