{"application": "software-mentions", "version": "0.9.0-SNAPSHOT", "revision": "54efbc6", "date": "2026-10-14T14:49+0000", "md5": "E7CD658F9A12D440ED7C9D23C2B5C7A3", "biblio": {"title": "A Survey on Multimodal Federated Learning", "authors": "Silvia Grosso, Nawel Benarba, Sara Bouchenak, Roberto Esposito, Mirko Polato"}, "mentions": [{"type": "software", "software-type": "software", "software-name": {"rawForm": "FedHAR", "normalizedForm": "FedHAR", "offsetStart": 0, "offsetEnd": 6}, "context": "FedHAR addresses label scarcity in human activity recognition by leveraging large amounts of unlabeled real-time sensor data .", "mentionContextAttributes": {"used": {"value": false, "score": 0.0001119375228881836}, "created": {"value": false, "score": 0.000264585018157959}, "shared": {"value": false, "score": 1.1920928955078125e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.000264585018157959}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[85]", "normalizedForm": "[85]", "refKey": 85}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "AimNet", "normalizedForm": "AimNet", "offsetStart": 0, "offsetEnd": 6}, "context": "AimNet, one of the first MMFL systems in the literature, fuses locally extracted visual and textual features through an aligning, integrating, and mapping network on the server to obtain fine-grained representations for vision-and-language problems.", "mentionContextAttributes": {"used": {"value": false, "score": 2.6941299438476562e-05}, "created": {"value": false, "score": 0.07612228393554688}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.07612228393554688}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "references": [{"label": "[44]", "normalizedForm": "[44]", "refKey": 44}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedSea", "normalizedForm": "FedSea", "offsetStart": 0, "offsetEnd": 6}, "context": "FedSea, instead, focuses on selectively aligning non-IID client features at the server to increase data homogeneity and mitigate client drift.", "mentionContextAttributes": {"used": {"value": false, "score": 6.568431854248047e-05}, "created": {"value": false, "score": 0.004849433898925781}, "shared": {"value": false, "score": 5.960464477539062e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.027053534984588623}, "shared": {"value": false, "score": 5.960464477539062e-07}}, "references": [{"label": "[78]", "normalizedForm": "[78]", "refKey": 78}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedHKD", "normalizedForm": "FedHKD", "offsetStart": 0, "offsetEnd": 6}, "context": "FedHKD instead follows a hierarchical knowledge disentanglement at both the modality and client levels , first learning modality-agnostic knowledge in a shared base model before fine-tuning modality-specific components, ensuring efficient adaptation with minimal updates.", "mentionContextAttributes": {"used": {"value": false, "score": 9.524822235107422e-05}, "created": {"value": false, "score": 1.71661376953125e-05}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.007363557815551758}, "created": {"value": false, "score": 6.496906280517578e-05}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "references": [{"label": "[134]", "normalizedForm": "[134]", "refKey": 134}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "AutoFed", "normalizedForm": "AutoFed", "offsetStart": 0, "offsetEnd": 7}, "context": "AutoFed and FLASH propose multimodal frameworks for autonomous driving systems that incorporate local cross-modal learning and model selection strategies during the aggregation phase .", "mentionContextAttributes": {"used": {"value": false, "score": 7.30752944946289e-05}, "created": {"value": false, "score": 0.2717921733856201}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.42818617820739746}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[74]", "normalizedForm": "[74]", "refKey": 74}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "AutoFed", "normalizedForm": "AutoFed", "offsetStart": 0, "offsetEnd": 7}, "context": "AutoFed adopts a strategic server-side client selection mechanism to manage environmental heterogeneity by filtering out divergent models and selecting mutually reinforcing clients.", "mentionContextAttributes": {"used": {"value": false, "score": 5.4001808166503906e-05}, "created": {"value": false, "score": 0.42818617820739746}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.42818617820739746}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[74]", "normalizedForm": "[74]", "refKey": 74}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "CreamFL", "normalizedForm": "CreamFL", "offsetStart": 0, "offsetEnd": 7}, "context": "CreamFL is the first MMFL framework effectively handling heterogeneity in local modalities, architectures, and tasks while only communicating knowledge derived from a public dataset .", "mentionContextAttributes": {"used": {"value": false, "score": 6.461143493652344e-05}, "created": {"value": false, "score": 0.017334938049316406}, "shared": {"value": false, "score": 1.1920928955078125e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 9.226799011230469e-05}, "created": {"value": false, "score": 0.017334938049316406}, "shared": {"value": false, "score": 1.1920928955078125e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "LG-FedAvg", "normalizedForm": "LG-FedAvg", "offsetStart": 0, "offsetEnd": 9}, "context": "LG-FedAvg introduces a hybrid architecture with personalized local layers for learning clientspecific representations, while shared base layers, hosted on the server, handle classification .", "mentionContextAttributes": {"used": {"value": false, "score": 3.552436828613281e-05}, "created": {"value": false, "score": 0.002541482448577881}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 3.552436828613281e-05}, "created": {"value": false, "score": 0.002541482448577881}, "shared": {"value": false, "score": 2.384185791015625e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedCMR", "normalizedForm": "FedCMR", "offsetStart": 0, "offsetEnd": 11}, "context": "FedCMR learns a shared subspace for cross-modal retrieval, reducing communication by aggregating subspace updates rather than full models.", "mentionContextAttributes": {"used": {"value": false, "score": 6.35385513305664e-05}, "created": {"value": false, "score": 1.3589859008789062e-05}, "shared": {"value": false, "score": 1.1920928955078125e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.3589859008789062e-05}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[92]", "normalizedForm": "[92]", "refKey": 92}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedSCMR", "normalizedForm": "FedSCMR", "offsetStart": 0, "offsetEnd": 12}, "context": "FedSCMR extends this approach by introducing multiple concurrent subspaces and an adaptive aggregation scheme, which enhances performance in highly heterogeneous environments.", "mentionContextAttributes": {"used": {"value": false, "score": 5.054473876953125e-05}, "created": {"value": true, "score": 0.9100821018218994}, "shared": {"value": false, "score": 1.1920928955078125e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": true, "score": 0.9100821018218994}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[93]", "normalizedForm": "[93]", "refKey": 93}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedIoT", "normalizedForm": "FedIoT", "offsetStart": 3, "offsetEnd": 9}, "context": "In FedIoT, thanks to an auxiliary labeled dataset on the server, classifiers trained on labeled data on the server from one modality can achieve decent performance on testing data from other modalities.", "mentionContextAttributes": {"used": {"value": false, "score": 0.00031554698944091797}, "created": {"value": false, "score": 0.0001818537712097168}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.00031554698944091797}, "created": {"value": false, "score": 0.0001818537712097168}, "shared": {"value": false, "score": 2.384185791015625e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "DisentAFL", "normalizedForm": "DisentAFL", "offsetStart": 3, "offsetEnd": 12}, "context": "In DisentAFL [53] 2024 (4) images/audio fully non-IID FedHKD [134] 2024 (4) images/audio fully non-IID addition, FedHCD does not require clients to possess all modalities but can operate effectively with only unimodal clients.", "mentionContextAttributes": {"used": {"value": false, "score": 0.007363557815551758}, "created": {"value": false, "score": 1.1682510375976562e-05}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.007363557815551758}, "created": {"value": false, "score": 6.496906280517578e-05}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "references": [{"label": "[53]", "normalizedForm": "[53]", "refKey": 53, "offsetStart": 64649, "offsetEnd": 64653}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedUSL", "normalizedForm": "FedUSL", "offsetStart": 3, "offsetEnd": 13}, "context": "In FedUSL , the authors propose a federated annotation method for multimodal sensing data in driving fatigue detection, leveraging four modalities simultaneously to capture correlations and complementarity with low computational complexity.", "mentionContextAttributes": {"used": {"value": false, "score": 6.961822509765625e-05}, "created": {"value": true, "score": 0.7006048560142517}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": true, "score": 0.7006048560142517}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[65]", "normalizedForm": "[65]", "refKey": 65}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "AimNet", "normalizedForm": "AimNet", "offsetStart": 5, "offsetEnd": 11}, "context": "Both AimNet and FedSea adopt server-side strategies for transforming data representations in the feature space .", "mentionContextAttributes": {"used": {"value": false, "score": 6.723403930664062e-05}, "created": {"value": false, "score": 0.027053534984588623}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.07612228393554688}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "references": [{"label": "[44]", "normalizedForm": "[44]", "refKey": 44}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "CreamFL", "normalizedForm": "CreamFL", "offsetStart": 7, "offsetEnd": 14}, "context": "Unlike CreamFL, the authors of FedMEKT consider a semi-supervised learning context that exploits the joint multimodal knowledge of the unlabeled data of participating clients using proxy data to obtain a generalized encoder [45].", "mentionContextAttributes": {"used": {"value": false, "score": 9.226799011230469e-05}, "created": {"value": false, "score": 4.863739013671875e-05}, "shared": {"value": false, "score": 1.1920928955078125e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 9.226799011230469e-05}, "created": {"value": false, "score": 0.017334938049316406}, "shared": {"value": false, "score": 1.1920928955078125e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "DisentAFL", "normalizedForm": "DisentAFL", "offsetStart": 7, "offsetEnd": 16}, "context": "Unlike DisentAFL, which relies on a centralized supernetwork for aggregation, FedHKD integrates a heterogeneity-aware aggregation strategy and a correlation-based fusion model to refine local models while dynamically handling diverse client updates.", "mentionContextAttributes": {"used": {"value": false, "score": 8.392333984375e-05}, "created": {"value": false, "score": 6.496906280517578e-05}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.007363557815551758}, "created": {"value": false, "score": 6.496906280517578e-05}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "references": [{"label": "[53]", "normalizedForm": "[53]", "refKey": 53}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedMSplit", "normalizedForm": "FedMSplit", "offsetStart": 9, "offsetEnd": 23}, "context": "Finally, FedMSplit uses a dynamic, multiview graph structure to capture correlations among clients with multi-task objectives.", "mentionContextAttributes": {"used": {"value": false, "score": 4.57763671875e-05}, "created": {"value": false, "score": 7.021427154541016e-05}, "shared": {"value": false, "score": 1.1920928955078125e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 4.57763671875e-05}, "created": {"value": false, "score": 7.021427154541016e-05}, "shared": {"value": false, "score": 1.1920928955078125e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FLASH", "normalizedForm": "FLASH", "offsetStart": 12, "offsetEnd": 17}, "context": "AutoFed and FLASH propose multimodal frameworks for autonomous driving systems that incorporate local cross-modal learning and model selection strategies during the aggregation phase .", "mentionContextAttributes": {"used": {"value": false, "score": 7.30752944946289e-05}, "created": {"value": false, "score": 0.2717921733856201}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.2717921733856201}, "shared": {"value": false, "score": 1.5497207641601562e-06}}, "references": [{"label": "[77]", "normalizedForm": "[77]", "refKey": 77}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FLASH", "normalizedForm": "FLASH", "offsetStart": 12, "offsetEnd": 17}, "context": "Conversely, FLASH employs a stochastic branch selection technique that reduces communication costs and maintains high performance, even under very short connection windows between vehicles and mobile edge computing (MEC) servers.", "mentionContextAttributes": {"used": {"value": false, "score": 2.5033950805664062e-05}, "created": {"value": false, "score": 0.0003744363784790039}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.2717921733856201}, "shared": {"value": false, "score": 1.5497207641601562e-06}}, "references": [{"label": "[77]", "normalizedForm": "[77]", "refKey": 77}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "Fed-inMM", "normalizedForm": "Fed-inMM", "offsetStart": 12, "offsetEnd": 24}, "context": "Finally, in Fed-inMM , the use of a hybrid fusion approach ensures more robust multimodal representation learning, effectively mitigating the impact of missing modalities.", "mentionContextAttributes": {"used": {"value": false, "score": 4.7087669372558594e-05}, "created": {"value": false, "score": 0.0012094378471374512}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 4.7087669372558594e-05}, "created": {"value": false, "score": 0.0012094378471374512}, "shared": {"value": false, "score": 2.384185791015625e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "Dis-entAFL", "normalizedForm": "Dis-entAFL", "offsetStart": 13, "offsetEnd": 23}, "context": "For example, Dis-entAFL enables clients to disentangle their learned representations into modality-specific and taskspecific components , which are uploaded to a server-hosted supernetwork that acts as a structured global repository.", "mentionContextAttributes": {"used": {"value": false, "score": 7.259845733642578e-05}, "created": {"value": false, "score": 0.003503262996673584}, "shared": {"value": false, "score": 7.62939453125e-06}}, "documentContextAttributes": {"used": {"value": false, "score": 7.259845733642578e-05}, "created": {"value": false, "score": 0.003503262996673584}, "shared": {"value": false, "score": 7.62939453125e-06}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedSea", "normalizedForm": "FedSea", "offsetStart": 16, "offsetEnd": 22}, "context": "Both AimNet and FedSea adopt server-side strategies for transforming data representations in the feature space .", "mentionContextAttributes": {"used": {"value": false, "score": 6.723403930664062e-05}, "created": {"value": false, "score": 0.027053534984588623}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.027053534984588623}, "shared": {"value": false, "score": 5.960464477539062e-07}}, "references": [{"label": "[78]", "normalizedForm": "[78]", "refKey": 78}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedAvg", "normalizedForm": "FedAvg", "offsetStart": 19, "offsetEnd": 25}, "context": "In this benchmark, FedAvg, FedProx, FedRS, SCAFFOLD, and FedOpt are evaluated with concatenation-and attention-based fusion under non-IID scenarios, showing that FedOpt generally outperforms other approaches, with some exceptions in low-heterogeneity cases .", "mentionContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": false, "score": 2.5033950805664062e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[101]", "normalizedForm": "[101]", "refKey": 101}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedProx", "normalizedForm": "FedProx", "offsetStart": 27, "offsetEnd": 34}, "context": "In this benchmark, FedAvg, FedProx, FedRS, SCAFFOLD, and FedOpt are evaluated with concatenation-and attention-based fusion under non-IID scenarios, showing that FedOpt generally outperforms other approaches, with some exceptions in low-heterogeneity cases .", "mentionContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": false, "score": 2.5033950805664062e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "Assistant", "normalizedForm": "Assistant", "offsetStart": 35, "offsetEnd": 44}, "context": "A real-world example is the Google Assistant ecosystem, where devices differ in available sensors and computational resources: Nest Hubs can process voice and visual signals entirely on-device to enhance interaction; Nest Minis handle only voice with a lightweight speech model; and Pixel phones process voice and text commands exploiting more complex model architectures .", "mentionContextAttributes": {"used": {"value": false, "score": 6.914138793945312e-05}, "created": {"value": false, "score": 0.004477262496948242}, "shared": {"value": false, "score": 2.6226043701171875e-06}}, "documentContextAttributes": {"used": {"value": false, "score": 6.914138793945312e-05}, "created": {"value": false, "score": 0.004477262496948242}, "shared": {"value": false, "score": 2.6226043701171875e-06}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedRS", "normalizedForm": "FedRS", "offsetStart": 36, "offsetEnd": 41}, "context": "In this benchmark, FedAvg, FedProx, FedRS, SCAFFOLD, and FedOpt are evaluated with concatenation-and attention-based fusion under non-IID scenarios, showing that FedOpt generally outperforms other approaches, with some exceptions in low-heterogeneity cases .", "mentionContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": false, "score": 2.5033950805664062e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": false, "score": 2.5033950805664062e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "SCAFFOLD", "normalizedForm": "SCAFFOLD", "offsetStart": 43, "offsetEnd": 51}, "context": "In this benchmark, FedAvg, FedProx, FedRS, SCAFFOLD, and FedOpt are evaluated with concatenation-and attention-based fusion under non-IID scenarios, showing that FedOpt generally outperforms other approaches, with some exceptions in low-heterogeneity cases .", "mentionContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": false, "score": 2.5033950805664062e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": false, "score": 2.5033950805664062e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedHKD", "normalizedForm": "FedHKD", "offsetStart": 54, "offsetEnd": 60}, "context": "In DisentAFL [53] 2024 (4) images/audio fully non-IID FedHKD [134] 2024 (4) images/audio fully non-IID addition, FedHCD does not require clients to possess all modalities but can operate effectively with only unimodal clients.", "mentionContextAttributes": {"used": {"value": false, "score": 0.007363557815551758}, "created": {"value": false, "score": 1.1682510375976562e-05}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.007363557815551758}, "created": {"value": false, "score": 6.496906280517578e-05}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "references": [{"label": "[134]", "normalizedForm": "[134]", "refKey": 134, "offsetStart": 64697, "offsetEnd": 64702}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedOpt", "normalizedForm": "FedOpt", "offsetStart": 57, "offsetEnd": 63}, "context": "In this benchmark, FedAvg, FedProx, FedRS, SCAFFOLD, and FedOpt are evaluated with concatenation-and attention-based fusion under non-IID scenarios, showing that FedOpt generally outperforms other approaches, with some exceptions in low-heterogeneity cases .", "mentionContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": false, "score": 2.5033950805664062e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedMobile", "normalizedForm": "FedMobile", "offsetStart": 61, "offsetEnd": 70}, "context": "Among the homogeneous MMFL systems exploiting KD , FedMobile introduces a knowledge contribution-aware multimodal FL framework that reconstructs missing features by distilling cross-node multimodal knowledge .", "mentionContextAttributes": {"used": {"value": false, "score": 3.62396240234375e-05}, "created": {"value": false, "score": 0.16788983345031738}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.16788983345031738}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[102]", "normalizedForm": "[102]", "refKey": 102}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedAvg", "normalizedForm": "FedAvg", "offsetStart": 63, "offsetEnd": 69}, "context": "A widely used aggregation strategy is the Federated Averaging (FedAvg) algorithm, which computes a weighted average of the local updates, as outlined in Eq. ( 1).", "mentionContextAttributes": {"used": {"value": false, "score": 0.00016951560974121094}, "created": {"value": false, "score": 0.0006458163261413574}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[101]", "normalizedForm": "[101]", "refKey": 101}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "IEEE Xplore", "normalizedForm": "IEEE Xplore", "offsetStart": 67, "offsetEnd": 78}, "context": "We searched major academic databases, such as ACM Digital Library, IEEE Xplore, Google Scholar, and DBLP, considering English-language publications up to early 2025.", "mentionContextAttributes": {"used": {"value": true, "score": 0.994820237159729}, "created": {"value": false, "score": 0.0008348226547241211}, "shared": {"value": false, "score": 1.8835067749023438e-05}}, "documentContextAttributes": {"used": {"value": true, "score": 0.994820237159729}, "created": {"value": false, "score": 0.0008348226547241211}, "shared": {"value": false, "score": 1.8835067749023438e-05}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedHKD", "normalizedForm": "FedHKD", "offsetStart": 78, "offsetEnd": 84}, "context": "Unlike DisentAFL, which relies on a centralized supernetwork for aggregation, FedHKD integrates a heterogeneity-aware aggregation strategy and a correlation-based fusion model to refine local models while dynamically handling diverse client updates.", "mentionContextAttributes": {"used": {"value": false, "score": 8.392333984375e-05}, "created": {"value": false, "score": 6.496906280517578e-05}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.007363557815551758}, "created": {"value": false, "score": 6.496906280517578e-05}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "references": [{"label": "[134]", "normalizedForm": "[134]", "refKey": 134}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "Google Scholar", "normalizedForm": "Google Scholar", "offsetStart": 80, "offsetEnd": 94}, "context": "We searched major academic databases, such as ACM Digital Library, IEEE Xplore, Google Scholar, and DBLP, considering English-language publications up to early 2025.", "mentionContextAttributes": {"used": {"value": true, "score": 0.994820237159729}, "created": {"value": false, "score": 0.0008347034454345703}, "shared": {"value": false, "score": 1.8835067749023438e-05}}, "documentContextAttributes": {"used": {"value": true, "score": 0.994820237159729}, "created": {"value": false, "score": 0.0008347034454345703}, "shared": {"value": false, "score": 1.8835067749023438e-05}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedHAR", "normalizedForm": "FedHAR", "offsetStart": 112, "offsetEnd": 118}, "context": "Among the homogeneous systems in which clients exploit hybrid FedExam [80] 2021 (2) image/tabular fully non-IID FedHAR [85] 2021 (2-3) time series partly non-IID FedInMM [64] 2023 (2-3) signal fully non-IID FedUSL [65] 2024 (4) video/signal partly non-IID FLPCS-MRF [86] 2025 (2) time series/image fully non-IID PSO-ACO [87] 2025 (5) image/signal,audio fully non-IID coordination Chen & Pan [88] 2024 (2) image/text fully -FedCMD [67] 2024 (2) video/signal partly IID/non-IID FedDAT [89] 2024 (2) image/text fully non-IID FedMed-ATL [68] 2022 (3) image not annotated non-IID FedPIA [90] 2025 (2) image/text fully non-IID Saeed et al. [75] 2020 (3) signal not annotated IID Client-Server cross-modal learning AutoFed [74] 2023 (2) signal partly non-IID CACMRN [73] 2023 (2) video/optical flow fully non-IID FedAA [91] 2023 (2) image/text fully non-IID FedCMR [92] 2021 (2) image/text fully -FedSCMR [93] 2024 (2) image/text fully -FLASH [77] 2022 (3) image/signal fully IID PFL-MCL [94] 2024 (3) image/text/audio,tabular fully non-IID prototype-based learning Buffalo [95] 2024 (2) image/text partly IID/non-IID MFCPL [96] 2025 (3) text/video/audio,image/text,signal fully non-IID PEPFCH [97] 2024 (2) image/text fully IID/non-IID PFMH [98] 2025 (2) image/text fully non-IID PT-FUCH [99] 2023 (2) image/text not annotated non-IID representation learning AimNet [44] 2020 (2) image/text not annotated non-IID FedMVP [100] 2024 (2) image, text fully IID/non-IID FedSea [78] 2023 (2-3) tabular,signal fully non-IID LG-FedAvg [101] 2020 (2) image/text fully non-IID knoweldge distillation FedCSR [9] 2023 (2) text/video fully non-IID FedMobile [102] 2025 (3) signal,audio/signal/image,image/signal fully non-IID HPFL [71] 2024 (2-3) image,audio/text/video fully IID/non-IID fusion techniques for local combination of multimodal information, FedExam [80] proposes a framework for predicting clinical outcomes in patients with COVID-19.", "mentionContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.1920928955078125e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.000264585018157959}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[85]", "normalizedForm": "[85]", "refKey": 85, "offsetStart": 38519, "offsetEnd": 38523}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedOpt", "normalizedForm": "FedOpt", "offsetStart": 162, "offsetEnd": 168}, "context": "In this benchmark, FedAvg, FedProx, FedRS, SCAFFOLD, and FedOpt are evaluated with concatenation-and attention-based fusion under non-IID scenarios, showing that FedOpt generally outperforms other approaches, with some exceptions in low-heterogeneity cases .", "mentionContextAttributes": {"used": {"value": false, "score": 0.19560211896896362}, "created": {"value": false, "score": 2.5033950805664062e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedAvg", "normalizedForm": "FedAvg", "offsetStart": 199, "offsetEnd": 205}, "context": "In the previous section, we have seen that many homogeneous MMFL systems propose local multimodal fusion or coordination techniques while keeping standard server-side aggregation algorithms, such as FedAvg, FedProx, and FedOpt .", "mentionContextAttributes": {"used": {"value": false, "score": 5.4001808166503906e-05}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[101]", "normalizedForm": "[101]", "refKey": 101}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedUSL", "normalizedForm": "FedUSL", "offsetStart": 207, "offsetEnd": 213}, "context": "Among the homogeneous systems in which clients exploit hybrid FedExam [80] 2021 (2) image/tabular fully non-IID FedHAR [85] 2021 (2-3) time series partly non-IID FedInMM [64] 2023 (2-3) signal fully non-IID FedUSL [65] 2024 (4) video/signal partly non-IID FLPCS-MRF [86] 2025 (2) time series/image fully non-IID PSO-ACO [87] 2025 (5) image/signal,audio fully non-IID coordination Chen & Pan [88] 2024 (2) image/text fully -FedCMD [67] 2024 (2) video/signal partly IID/non-IID FedDAT [89] 2024 (2) image/text fully non-IID FedMed-ATL [68] 2022 (3) image not annotated non-IID FedPIA [90] 2025 (2) image/text fully non-IID Saeed et al. [75] 2020 (3) signal not annotated IID Client-Server cross-modal learning AutoFed [74] 2023 (2) signal partly non-IID CACMRN [73] 2023 (2) video/optical flow fully non-IID FedAA [91] 2023 (2) image/text fully non-IID FedCMR [92] 2021 (2) image/text fully -FedSCMR [93] 2024 (2) image/text fully -FLASH [77] 2022 (3) image/signal fully IID PFL-MCL [94] 2024 (3) image/text/audio,tabular fully non-IID prototype-based learning Buffalo [95] 2024 (2) image/text partly IID/non-IID MFCPL [96] 2025 (3) text/video/audio,image/text,signal fully non-IID PEPFCH [97] 2024 (2) image/text fully IID/non-IID PFMH [98] 2025 (2) image/text fully non-IID PT-FUCH [99] 2023 (2) image/text not annotated non-IID representation learning AimNet [44] 2020 (2) image/text not annotated non-IID FedMVP [100] 2024 (2) image, text fully IID/non-IID FedSea [78] 2023 (2-3) tabular,signal fully non-IID LG-FedAvg [101] 2020 (2) image/text fully non-IID knoweldge distillation FedCSR [9] 2023 (2) text/video fully non-IID FedMobile [102] 2025 (3) signal,audio/signal/image,image/signal fully non-IID HPFL [71] 2024 (2-3) image,audio/text/video fully IID/non-IID fusion techniques for local combination of multimodal information, FedExam [80] proposes a framework for predicting clinical outcomes in patients with COVID-19.", "mentionContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.1920928955078125e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": true, "score": 0.7006048560142517}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[65]", "normalizedForm": "[65]", "refKey": 65, "offsetStart": 38614, "offsetEnd": 38618}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedProx", "normalizedForm": "FedProx", "offsetStart": 207, "offsetEnd": 214}, "context": "In the previous section, we have seen that many homogeneous MMFL systems propose local multimodal fusion or coordination techniques while keeping standard server-side aggregation algorithms, such as FedAvg, FedProx, and FedOpt .", "mentionContextAttributes": {"used": {"value": false, "score": 5.4001808166503906e-05}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedOpt", "normalizedForm": "FedOpt", "offsetStart": 220, "offsetEnd": 237}, "context": "In the previous section, we have seen that many homogeneous MMFL systems propose local multimodal fusion or coordination techniques while keeping standard server-side aggregation algorithms, such as FedAvg, FedProx, and FedOpt .", "mentionContextAttributes": {"used": {"value": false, "score": 5.4001808166503906e-05}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": false, "score": 0.1956024169921875}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "AutoFed", "normalizedForm": "AutoFed", "offsetStart": 708, "offsetEnd": 715}, "context": "Among the homogeneous systems in which clients exploit hybrid FedExam [80] 2021 (2) image/tabular fully non-IID FedHAR [85] 2021 (2-3) time series partly non-IID FedInMM [64] 2023 (2-3) signal fully non-IID FedUSL [65] 2024 (4) video/signal partly non-IID FLPCS-MRF [86] 2025 (2) time series/image fully non-IID PSO-ACO [87] 2025 (5) image/signal,audio fully non-IID coordination Chen & Pan [88] 2024 (2) image/text fully -FedCMD [67] 2024 (2) video/signal partly IID/non-IID FedDAT [89] 2024 (2) image/text fully non-IID FedMed-ATL [68] 2022 (3) image not annotated non-IID FedPIA [90] 2025 (2) image/text fully non-IID Saeed et al. [75] 2020 (3) signal not annotated IID Client-Server cross-modal learning AutoFed [74] 2023 (2) signal partly non-IID CACMRN [73] 2023 (2) video/optical flow fully non-IID FedAA [91] 2023 (2) image/text fully non-IID FedCMR [92] 2021 (2) image/text fully -FedSCMR [93] 2024 (2) image/text fully -FLASH [77] 2022 (3) image/signal fully IID PFL-MCL [94] 2024 (3) image/text/audio,tabular fully non-IID prototype-based learning Buffalo [95] 2024 (2) image/text partly IID/non-IID MFCPL [96] 2025 (3) text/video/audio,image/text,signal fully non-IID PEPFCH [97] 2024 (2) image/text fully IID/non-IID PFMH [98] 2025 (2) image/text fully non-IID PT-FUCH [99] 2023 (2) image/text not annotated non-IID representation learning AimNet [44] 2020 (2) image/text not annotated non-IID FedMVP [100] 2024 (2) image, text fully IID/non-IID FedSea [78] 2023 (2-3) tabular,signal fully non-IID LG-FedAvg [101] 2020 (2) image/text fully non-IID knoweldge distillation FedCSR [9] 2023 (2) text/video fully non-IID FedMobile [102] 2025 (3) signal,audio/signal/image,image/signal fully non-IID HPFL [71] 2024 (2-3) image,audio/text/video fully IID/non-IID fusion techniques for local combination of multimodal information, FedExam [80] proposes a framework for predicting clinical outcomes in patients with COVID-19.", "mentionContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.1920928955078125e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.42818617820739746}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[74]", "normalizedForm": "[74]", "refKey": 74, "offsetStart": 39116, "offsetEnd": 39120}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedCMR", "normalizedForm": "FedCMR", "offsetStart": 851, "offsetEnd": 857}, "context": "Among the homogeneous systems in which clients exploit hybrid FedExam [80] 2021 (2) image/tabular fully non-IID FedHAR [85] 2021 (2-3) time series partly non-IID FedInMM [64] 2023 (2-3) signal fully non-IID FedUSL [65] 2024 (4) video/signal partly non-IID FLPCS-MRF [86] 2025 (2) time series/image fully non-IID PSO-ACO [87] 2025 (5) image/signal,audio fully non-IID coordination Chen & Pan [88] 2024 (2) image/text fully -FedCMD [67] 2024 (2) video/signal partly IID/non-IID FedDAT [89] 2024 (2) image/text fully non-IID FedMed-ATL [68] 2022 (3) image not annotated non-IID FedPIA [90] 2025 (2) image/text fully non-IID Saeed et al. [75] 2020 (3) signal not annotated IID Client-Server cross-modal learning AutoFed [74] 2023 (2) signal partly non-IID CACMRN [73] 2023 (2) video/optical flow fully non-IID FedAA [91] 2023 (2) image/text fully non-IID FedCMR [92] 2021 (2) image/text fully -FedSCMR [93] 2024 (2) image/text fully -FLASH [77] 2022 (3) image/signal fully IID PFL-MCL [94] 2024 (3) image/text/audio,tabular fully non-IID prototype-based learning Buffalo [95] 2024 (2) image/text partly IID/non-IID MFCPL [96] 2025 (3) text/video/audio,image/text,signal fully non-IID PEPFCH [97] 2024 (2) image/text fully IID/non-IID PFMH [98] 2025 (2) image/text fully non-IID PT-FUCH [99] 2023 (2) image/text not annotated non-IID representation learning AimNet [44] 2020 (2) image/text not annotated non-IID FedMVP [100] 2024 (2) image, text fully IID/non-IID FedSea [78] 2023 (2-3) tabular,signal fully non-IID LG-FedAvg [101] 2020 (2) image/text fully non-IID knoweldge distillation FedCSR [9] 2023 (2) text/video fully non-IID FedMobile [102] 2025 (3) signal,audio/signal/image,image/signal fully non-IID HPFL [71] 2024 (2-3) image,audio/text/video fully IID/non-IID fusion techniques for local combination of multimodal information, FedExam [80] proposes a framework for predicting clinical outcomes in patients with COVID-19.", "mentionContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.1920928955078125e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.3589859008789062e-05}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[92]", "normalizedForm": "[92]", "refKey": 92, "offsetStart": 39258, "offsetEnd": 39262}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedSCMR", "normalizedForm": "FedSCMR", "offsetStart": 890, "offsetEnd": 897}, "context": "Among the homogeneous systems in which clients exploit hybrid FedExam [80] 2021 (2) image/tabular fully non-IID FedHAR [85] 2021 (2-3) time series partly non-IID FedInMM [64] 2023 (2-3) signal fully non-IID FedUSL [65] 2024 (4) video/signal partly non-IID FLPCS-MRF [86] 2025 (2) time series/image fully non-IID PSO-ACO [87] 2025 (5) image/signal,audio fully non-IID coordination Chen & Pan [88] 2024 (2) image/text fully -FedCMD [67] 2024 (2) video/signal partly IID/non-IID FedDAT [89] 2024 (2) image/text fully non-IID FedMed-ATL [68] 2022 (3) image not annotated non-IID FedPIA [90] 2025 (2) image/text fully non-IID Saeed et al. [75] 2020 (3) signal not annotated IID Client-Server cross-modal learning AutoFed [74] 2023 (2) signal partly non-IID CACMRN [73] 2023 (2) video/optical flow fully non-IID FedAA [91] 2023 (2) image/text fully non-IID FedCMR [92] 2021 (2) image/text fully -FedSCMR [93] 2024 (2) image/text fully -FLASH [77] 2022 (3) image/signal fully IID PFL-MCL [94] 2024 (3) image/text/audio,tabular fully non-IID prototype-based learning Buffalo [95] 2024 (2) image/text partly IID/non-IID MFCPL [96] 2025 (3) text/video/audio,image/text,signal fully non-IID PEPFCH [97] 2024 (2) image/text fully IID/non-IID PFMH [98] 2025 (2) image/text fully non-IID PT-FUCH [99] 2023 (2) image/text not annotated non-IID representation learning AimNet [44] 2020 (2) image/text not annotated non-IID FedMVP [100] 2024 (2) image, text fully IID/non-IID FedSea [78] 2023 (2-3) tabular,signal fully non-IID LG-FedAvg [101] 2020 (2) image/text fully non-IID knoweldge distillation FedCSR [9] 2023 (2) text/video fully non-IID FedMobile [102] 2025 (3) signal,audio/signal/image,image/signal fully non-IID HPFL [71] 2024 (2-3) image,audio/text/video fully IID/non-IID fusion techniques for local combination of multimodal information, FedExam [80] proposes a framework for predicting clinical outcomes in patients with COVID-19.", "mentionContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.1920928955078125e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": true, "score": 0.9100821018218994}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[93]", "normalizedForm": "[93]", "refKey": 93, "offsetStart": 39298, "offsetEnd": 39302}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FLASH", "normalizedForm": "FLASH", "offsetStart": 930, "offsetEnd": 935}, "context": "Among the homogeneous systems in which clients exploit hybrid FedExam [80] 2021 (2) image/tabular fully non-IID FedHAR [85] 2021 (2-3) time series partly non-IID FedInMM [64] 2023 (2-3) signal fully non-IID FedUSL [65] 2024 (4) video/signal partly non-IID FLPCS-MRF [86] 2025 (2) time series/image fully non-IID PSO-ACO [87] 2025 (5) image/signal,audio fully non-IID coordination Chen & Pan [88] 2024 (2) image/text fully -FedCMD [67] 2024 (2) video/signal partly IID/non-IID FedDAT [89] 2024 (2) image/text fully non-IID FedMed-ATL [68] 2022 (3) image not annotated non-IID FedPIA [90] 2025 (2) image/text fully non-IID Saeed et al. [75] 2020 (3) signal not annotated IID Client-Server cross-modal learning AutoFed [74] 2023 (2) signal partly non-IID CACMRN [73] 2023 (2) video/optical flow fully non-IID FedAA [91] 2023 (2) image/text fully non-IID FedCMR [92] 2021 (2) image/text fully -FedSCMR [93] 2024 (2) image/text fully -FLASH [77] 2022 (3) image/signal fully IID PFL-MCL [94] 2024 (3) image/text/audio,tabular fully non-IID prototype-based learning Buffalo [95] 2024 (2) image/text partly IID/non-IID MFCPL [96] 2025 (3) text/video/audio,image/text,signal fully non-IID PEPFCH [97] 2024 (2) image/text fully IID/non-IID PFMH [98] 2025 (2) image/text fully non-IID PT-FUCH [99] 2023 (2) image/text not annotated non-IID representation learning AimNet [44] 2020 (2) image/text not annotated non-IID FedMVP [100] 2024 (2) image, text fully IID/non-IID FedSea [78] 2023 (2-3) tabular,signal fully non-IID LG-FedAvg [101] 2020 (2) image/text fully non-IID knoweldge distillation FedCSR [9] 2023 (2) text/video fully non-IID FedMobile [102] 2025 (3) signal,audio/signal/image,image/signal fully non-IID HPFL [71] 2024 (2-3) image,audio/text/video fully IID/non-IID fusion techniques for local combination of multimodal information, FedExam [80] proposes a framework for predicting clinical outcomes in patients with COVID-19.", "mentionContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.1920928955078125e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.2717921733856201}, "shared": {"value": false, "score": 1.5497207641601562e-06}}, "references": [{"label": "[77]", "normalizedForm": "[77]", "refKey": 77, "offsetStart": 39336, "offsetEnd": 39340}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "AimNet", "normalizedForm": "AimNet", "offsetStart": 1353, "offsetEnd": 1359}, "context": "Among the homogeneous systems in which clients exploit hybrid FedExam [80] 2021 (2) image/tabular fully non-IID FedHAR [85] 2021 (2-3) time series partly non-IID FedInMM [64] 2023 (2-3) signal fully non-IID FedUSL [65] 2024 (4) video/signal partly non-IID FLPCS-MRF [86] 2025 (2) time series/image fully non-IID PSO-ACO [87] 2025 (5) image/signal,audio fully non-IID coordination Chen & Pan [88] 2024 (2) image/text fully -FedCMD [67] 2024 (2) video/signal partly IID/non-IID FedDAT [89] 2024 (2) image/text fully non-IID FedMed-ATL [68] 2022 (3) image not annotated non-IID FedPIA [90] 2025 (2) image/text fully non-IID Saeed et al. [75] 2020 (3) signal not annotated IID Client-Server cross-modal learning AutoFed [74] 2023 (2) signal partly non-IID CACMRN [73] 2023 (2) video/optical flow fully non-IID FedAA [91] 2023 (2) image/text fully non-IID FedCMR [92] 2021 (2) image/text fully -FedSCMR [93] 2024 (2) image/text fully -FLASH [77] 2022 (3) image/signal fully IID PFL-MCL [94] 2024 (3) image/text/audio,tabular fully non-IID prototype-based learning Buffalo [95] 2024 (2) image/text partly IID/non-IID MFCPL [96] 2025 (3) text/video/audio,image/text,signal fully non-IID PEPFCH [97] 2024 (2) image/text fully IID/non-IID PFMH [98] 2025 (2) image/text fully non-IID PT-FUCH [99] 2023 (2) image/text not annotated non-IID representation learning AimNet [44] 2020 (2) image/text not annotated non-IID FedMVP [100] 2024 (2) image, text fully IID/non-IID FedSea [78] 2023 (2-3) tabular,signal fully non-IID LG-FedAvg [101] 2020 (2) image/text fully non-IID knoweldge distillation FedCSR [9] 2023 (2) text/video fully non-IID FedMobile [102] 2025 (3) signal,audio/signal/image,image/signal fully non-IID HPFL [71] 2024 (2-3) image,audio/text/video fully IID/non-IID fusion techniques for local combination of multimodal information, FedExam [80] proposes a framework for predicting clinical outcomes in patients with COVID-19.", "mentionContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.1920928955078125e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.07612228393554688}, "shared": {"value": false, "score": 4.76837158203125e-07}}, "references": [{"label": "[44]", "normalizedForm": "[44]", "refKey": 44, "offsetStart": 39760, "offsetEnd": 39764}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedSea", "normalizedForm": "FedSea", "offsetStart": 1459, "offsetEnd": 1465}, "context": "Among the homogeneous systems in which clients exploit hybrid FedExam [80] 2021 (2) image/tabular fully non-IID FedHAR [85] 2021 (2-3) time series partly non-IID FedInMM [64] 2023 (2-3) signal fully non-IID FedUSL [65] 2024 (4) video/signal partly non-IID FLPCS-MRF [86] 2025 (2) time series/image fully non-IID PSO-ACO [87] 2025 (5) image/signal,audio fully non-IID coordination Chen & Pan [88] 2024 (2) image/text fully -FedCMD [67] 2024 (2) video/signal partly IID/non-IID FedDAT [89] 2024 (2) image/text fully non-IID FedMed-ATL [68] 2022 (3) image not annotated non-IID FedPIA [90] 2025 (2) image/text fully non-IID Saeed et al. [75] 2020 (3) signal not annotated IID Client-Server cross-modal learning AutoFed [74] 2023 (2) signal partly non-IID CACMRN [73] 2023 (2) video/optical flow fully non-IID FedAA [91] 2023 (2) image/text fully non-IID FedCMR [92] 2021 (2) image/text fully -FedSCMR [93] 2024 (2) image/text fully -FLASH [77] 2022 (3) image/signal fully IID PFL-MCL [94] 2024 (3) image/text/audio,tabular fully non-IID prototype-based learning Buffalo [95] 2024 (2) image/text partly IID/non-IID MFCPL [96] 2025 (3) text/video/audio,image/text,signal fully non-IID PEPFCH [97] 2024 (2) image/text fully IID/non-IID PFMH [98] 2025 (2) image/text fully non-IID PT-FUCH [99] 2023 (2) image/text not annotated non-IID representation learning AimNet [44] 2020 (2) image/text not annotated non-IID FedMVP [100] 2024 (2) image, text fully IID/non-IID FedSea [78] 2023 (2-3) tabular,signal fully non-IID LG-FedAvg [101] 2020 (2) image/text fully non-IID knoweldge distillation FedCSR [9] 2023 (2) text/video fully non-IID FedMobile [102] 2025 (3) signal,audio/signal/image,image/signal fully non-IID HPFL [71] 2024 (2-3) image,audio/text/video fully IID/non-IID fusion techniques for local combination of multimodal information, FedExam [80] proposes a framework for predicting clinical outcomes in patients with COVID-19.", "mentionContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.1920928955078125e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.027053534984588623}, "shared": {"value": false, "score": 5.960464477539062e-07}}, "references": [{"label": "[78]", "normalizedForm": "[78]", "refKey": 78, "offsetStart": 39866, "offsetEnd": 39870}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedAvg", "normalizedForm": "FedAvg", "offsetStart": 1514, "offsetEnd": 1520}, "context": "Among the homogeneous systems in which clients exploit hybrid FedExam [80] 2021 (2) image/tabular fully non-IID FedHAR [85] 2021 (2-3) time series partly non-IID FedInMM [64] 2023 (2-3) signal fully non-IID FedUSL [65] 2024 (4) video/signal partly non-IID FLPCS-MRF [86] 2025 (2) time series/image fully non-IID PSO-ACO [87] 2025 (5) image/signal,audio fully non-IID coordination Chen & Pan [88] 2024 (2) image/text fully -FedCMD [67] 2024 (2) video/signal partly IID/non-IID FedDAT [89] 2024 (2) image/text fully non-IID FedMed-ATL [68] 2022 (3) image not annotated non-IID FedPIA [90] 2025 (2) image/text fully non-IID Saeed et al. [75] 2020 (3) signal not annotated IID Client-Server cross-modal learning AutoFed [74] 2023 (2) signal partly non-IID CACMRN [73] 2023 (2) video/optical flow fully non-IID FedAA [91] 2023 (2) image/text fully non-IID FedCMR [92] 2021 (2) image/text fully -FedSCMR [93] 2024 (2) image/text fully -FLASH [77] 2022 (3) image/signal fully IID PFL-MCL [94] 2024 (3) image/text/audio,tabular fully non-IID prototype-based learning Buffalo [95] 2024 (2) image/text partly IID/non-IID MFCPL [96] 2025 (3) text/video/audio,image/text,signal fully non-IID PEPFCH [97] 2024 (2) image/text fully IID/non-IID PFMH [98] 2025 (2) image/text fully non-IID PT-FUCH [99] 2023 (2) image/text not annotated non-IID representation learning AimNet [44] 2020 (2) image/text not annotated non-IID FedMVP [100] 2024 (2) image, text fully IID/non-IID FedSea [78] 2023 (2-3) tabular,signal fully non-IID LG-FedAvg [101] 2020 (2) image/text fully non-IID knoweldge distillation FedCSR [9] 2023 (2) text/video fully non-IID FedMobile [102] 2025 (3) signal,audio/signal/image,image/signal fully non-IID HPFL [71] 2024 (2-3) image,audio/text/video fully IID/non-IID fusion techniques for local combination of multimodal information, FedExam [80] proposes a framework for predicting clinical outcomes in patients with COVID-19.", "mentionContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.1920928955078125e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": true, "score": 0.9479824900627136}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[101]", "normalizedForm": "[101]", "refKey": 101, "offsetStart": 39921, "offsetEnd": 39926}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FedMobile", "normalizedForm": "FedMobile", "offsetStart": 1629, "offsetEnd": 1638}, "context": "Among the homogeneous systems in which clients exploit hybrid FedExam [80] 2021 (2) image/tabular fully non-IID FedHAR [85] 2021 (2-3) time series partly non-IID FedInMM [64] 2023 (2-3) signal fully non-IID FedUSL [65] 2024 (4) video/signal partly non-IID FLPCS-MRF [86] 2025 (2) time series/image fully non-IID PSO-ACO [87] 2025 (5) image/signal,audio fully non-IID coordination Chen & Pan [88] 2024 (2) image/text fully -FedCMD [67] 2024 (2) video/signal partly IID/non-IID FedDAT [89] 2024 (2) image/text fully non-IID FedMed-ATL [68] 2022 (3) image not annotated non-IID FedPIA [90] 2025 (2) image/text fully non-IID Saeed et al. [75] 2020 (3) signal not annotated IID Client-Server cross-modal learning AutoFed [74] 2023 (2) signal partly non-IID CACMRN [73] 2023 (2) video/optical flow fully non-IID FedAA [91] 2023 (2) image/text fully non-IID FedCMR [92] 2021 (2) image/text fully -FedSCMR [93] 2024 (2) image/text fully -FLASH [77] 2022 (3) image/signal fully IID PFL-MCL [94] 2024 (3) image/text/audio,tabular fully non-IID prototype-based learning Buffalo [95] 2024 (2) image/text partly IID/non-IID MFCPL [96] 2025 (3) text/video/audio,image/text,signal fully non-IID PEPFCH [97] 2024 (2) image/text fully IID/non-IID PFMH [98] 2025 (2) image/text fully non-IID PT-FUCH [99] 2023 (2) image/text not annotated non-IID representation learning AimNet [44] 2020 (2) image/text not annotated non-IID FedMVP [100] 2024 (2) image, text fully IID/non-IID FedSea [78] 2023 (2-3) tabular,signal fully non-IID LG-FedAvg [101] 2020 (2) image/text fully non-IID knoweldge distillation FedCSR [9] 2023 (2) text/video fully non-IID FedMobile [102] 2025 (3) signal,audio/signal/image,image/signal fully non-IID HPFL [71] 2024 (2-3) image,audio/text/video fully IID/non-IID fusion techniques for local combination of multimodal information, FedExam [80] proposes a framework for predicting clinical outcomes in patients with COVID-19.", "mentionContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 1.1920928955078125e-06}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.16788983345031738}, "shared": {"value": false, "score": 2.384185791015625e-07}}, "references": [{"label": "[102]", "normalizedForm": "[102]", "refKey": 102, "offsetStart": 40039, "offsetEnd": 40044}]}, {"type": "software", "software-type": "software", "software-name": {"rawForm": "FLASH", "normalizedForm": "FLASH", "offsetStart": 3284, "offsetEnd": 3289}, "context": "Application Domains References Datasets Models Tasks Healthcare [76], [88], [67], [80], [68], [48], [49], [75], [84], [87], [47], [90], [65], [69], [56], [60], [95], [115], [119], [121], [117], [114], [120] ISIC 2020, IU-XRay, BioVid Emo DB, IXI, BraTS 2020, BraTS 2021, MSWML 2022, ABCs Task 2, PTB-XL, MIMIC-III, Sleep-EDF, WESAD, BiAffect, LiTS17, 3D-IRCADb-01, Multi-Atlas, SLIVER07, CHAOS19, KORA, COVID-19 image data collection, POCUS, VQA-RAD, Slake-English, MIMIC-CXR, NIH Open-I, CheXpert, fastMRI, uMR, UP Fall RNN [47,49,70,115,121], CNN [47,60,65,67,69,76,88,114,121], MLP [76,114,119], GAN [49,68], Transformer [49,88,95], RFNet [79], BERT-based [115], U-Net [119,120], GloVe [70] melanoma detection [76], COVID-19 detection [69], brain tumor segmentation [56,79], liver segmentation [120], medical report generation [88,95], biomedical VQA [90,95], emotion recognition [67,75], depression detection [84], alzheimer detection [47,114], clinical outcome prediction [49,80], neuroimage reconstruction [68,117,119], driving fatigue detection [65], inhospital mortality prediction [121], classification [48,115], fall detection [60,87], sleep stage detection [75] Activity/Emotion recognition Epic-Kitchens, EGTEA Gaze+, Stanford-ECM, HARWE, DEAP, RealWorld, UCI-HAR, KU-HAR, AReM, MELD, CREMA-D, UCF101, MiT10, MiT51, HAR dataset, UP Fall, Multimodal Data, MMC-PCL-Activity, HHAR, MobiAct, WiFi-CSI, Multimodal-EA, Ego-Exo-AR, Kinetics, Finegym, UR Fall Detection, mHealth, Opportunity, ActionSense, IEMOCAP, CMU-MOSI, CMU-MOSEI, USC dataset, AMIGOS, DREAMER, AffectNet, Seed-V, MEAD, MM-Fi, ActionSense, MMHAR, IEMOCAP, MHEALTH, URFALL, Amazon Product Reviews, VSM, CREMA-D, UTD-MHAD, CMU-MOSI CNN [47,54,60,65,73,78,83,123,125], FFNN [61], RNN [45,47,54,55,64,71,81,83,94,124,128], attentionbased [62,83,85], AEs [45,55], Transformer [54,62,73,81,82,94,134], MLP [54,72,73,[111][112][113]128], Foundation model [53], Multimodal MBFM [126] activity recognition [45, 47, 48, 55, 64, 72, 73, 75, 78, 82, 83, 85, 111-113, 118, 122-124, 126-128, 134], emotion recognition [48,54,57,61,62,64,65,70,71,81,96,113,125,128] fall detection [60], user behavior prediction [94] Computer vision [54], [57], [116], [121], [125] ModelNet40, NYU-Depth-v2, AV-MNIST, Colored MNIST CNN [54,116,121,125], RNN [54,116,121], Transformer [54], MLP [54] 3D object recognition [57,121,125], digit classification [54,116,125], scene recognition [54] Vision-Language models [44], [92], [46], [9], [91], [89], [93], [99], [48], [100], [97], [98], [101], [116] Flicker30K, MSCOCO, VQA v2.0, MIR-Flickr25K, Wikipedia, Pascal Sentence, Chinese CS dataset, VizWiz, COCO QA, Art, GQA, Abstract, UR-Funny, Hateful-Memes, CrisisMMD, CUB-200, Oxford Flower, MIRFLICKR-25K, NUS-WIDE, FashionVC, Ssense CNN [9, 44, 46, 91-93, 99, 101, 116], RNN [9,46,48,91,101,116], Transformer [9,89], BERT-based [46,92,93,99], hypernetworks [97] image captioning [44,91], fine-grained image classification [100], cross-modal retrieval [46,92,93,[97][98][99], VE [89], VCR [89], NLVR [89], VQA [44,46,89,101], cued speech recognition [9], hate speech detection [48], crisis information detection [48], humor detection [116] Autonomous driving [74], [77], [57], [47], [71], [87], [110], [127] Oxford Radar RobotCar, nuScenes, FLASH, Vehicle Sensor, RGB-Thermal, Dataset [136], NYC Bike, NYC Taxi, Beijing (BJ) Taxi, Leddar PixSet, UrbanSound8K", "mentionContextAttributes": {"used": {"value": true, "score": 0.7789656519889832}, "created": {"value": false, "score": 6.4373016357421875e-06}, "shared": {"value": false, "score": 1.5497207641601562e-06}}, "documentContextAttributes": {"used": {"value": true, "score": 0.9436798691749573}, "created": {"value": false, "score": 0.2717921733856201}, "shared": {"value": false, "score": 1.5497207641601562e-06}}, "references": [{"label": "[77]", "normalizedForm": "[77]", "refKey": 77}]}], "references": [{"refKey": 85, "tei": "\n\t\n\t\tFedHAR: Semi-Supervised Online Learning for Personalized Federated Human Activity Recognition\n\t\t\n\t\t\tHongzhengYu\n\t\t\n\t\t\n\t\t\tZekaiChen\n\t\t\t0000-0002-5564-137X\n\t\t\n\t\t\n\t\t\tXiaoZhang\n\t\t\t0000-0003-0824-9284\n\t\t\n\t\t\n\t\t\tXuChen\n\t\t\t0000-0001-9943-6020\n\t\t\n\t\t\n\t\t\tFuzhenZhuang\n\t\t\t0000-0001-9170-7009\n\t\t\n\t\t\n\t\t\tHuiXiong\n\t\t\n\t\t\n\t\t\tXiuzhenCheng\n\t\t\t0000-0001-5912-4647\n\t\t\n\t\t10.1109/tmc.2021.3136853\n\t\n\t\n\t\tIEEE Transactions on Mobile Computing\n\t\tIEEE Trans. on Mobile Comput.\n\t\t1536-1233\n\t\t2161-9875\n\t\t\n\t\t\t22\n\t\t\t6\n\t\t\t\n\t\t\t2021\n\t\t\tInstitute of Electrical and Electronics Engineers (IEEE)\n\t\t\n\t\n\n"}, {"refKey": 44, "tei": "\n\t\n\t\tFederated Learning for Vision-and-Language Grounding Problems\n\t\t\n\t\t\tFenglinLiu\n\t\t\n\t\t\n\t\t\tXianWu\n\t\t\n\t\t\n\t\t\tShenGe\n\t\t\n\t\t\n\t\t\tWeiFan\n\t\t\n\t\t\n\t\t\tYuexianZou\n\t\t\n\t\t10.1609/aaai.v34i07.6824\n\t\n\t\n\t\tProceedings of the AAAI Conference on Artificial Intelligence\n\t\tAAAI\n\t\t2159-5399\n\t\t2374-3468\n\t\t\n\t\t\t34\n\t\t\t07\n\t\t\t\n\t\t\t2020\n\t\t\tAssociation for the Advancement of Artificial Intelligence (AAAI)\n\t\t\n\t\n\n"}, {"refKey": 78, "tei": "\n\t\n\t\tFedSea: Federated Learning via Selective Feature Alignment for Non-IID Multimodal Data\n\t\t\n\t\t\tMinTan\n\t\t\t0000-0002-1842-4050\n\t\t\n\t\t\n\t\t\tYinfuFeng\n\t\t\t0000-0001-9136-0965\n\t\t\n\t\t\n\t\t\tLingqiangChu\n\t\t\t0000-0003-0367-8397\n\t\t\n\t\t\n\t\t\tJingchengShi\n\t\t\t0000-0002-3688-4069\n\t\t\n\t\t\n\t\t\tRongXiao\n\t\t\t0000-0001-7793-6040\n\t\t\n\t\t\n\t\t\tHaihongTang\n\t\t\t0000-0001-6797-2988\n\t\t\n\t\t\n\t\t\tJunYu\n\t\t\t0000-0003-1922-7283\n\t\t\n\t\t10.1109/tmm.2023.3340109\n\t\n\t\n\t\tIEEE Transactions on Multimedia\n\t\tIEEE Trans. Multimedia\n\t\t1520-9210\n\t\t1941-0077\n\t\t\n\t\t\t26\n\t\t\t\n\t\t\t2023\n\t\t\tInstitute of Electrical and Electronics Engineers (IEEE)\n\t\t\n\t\n\n"}, {"refKey": 134, "tei": "\n\t\n\t\tTowards Efficient Heterogeneous Multi-Modal Federated Learning with Hierarchical Knowledge Disentanglement\n\t\t\n\t\t\tXingchenWang\n\t\t\t0000-0003-1352-7445\n\t\t\n\t\t\n\t\t\tHaoyuWang\n\t\t\t0000-0001-7485-6213\n\t\t\n\t\t\n\t\t\tFeijieWu\n\t\t\t0000-0003-0541-1901\n\t\t\n\t\t\n\t\t\tTianciLiu\n\t\t\t0000-0002-8396-8564\n\t\t\n\t\t\n\t\t\tQimingCao\n\t\t\t0000-0002-3329-3239\n\t\t\n\t\t\n\t\t\tLuSu\n\t\t\t0000-0001-7223-543X\n\t\t\n\t\t10.1145/3666025.3699360\n\t\n\t\n\t\tProceedings of the 22nd ACM Conference on Embedded Networked Sensor Systems\n\t\tthe 22nd ACM Conference on Embedded Networked Sensor Systems\n\t\t\n\t\t\tACM\n\t\t\t2024\n\t\t\t\n\t\t\n\t\n\n"}, {"refKey": 74, "tei": "\n\t\n\t\t\n\t\t\tTZheng\n\t\t\n\t\t\n\t\t\tALi\n\t\t\n\t\t\n\t\t\tZChen\n\t\t\n\t\t\n\t\t\tHWang\n\t\t\n\t\t\n\t\t\tJLuo\n\t\t\n\t\tarXiv preprint. 2023;2302.08646\n\t\tAut-oFed: Heterogeneity-Aware Federated Multimodal Learning for Robust Autonomous Driving\n\t\t\n\t\n\n"}, {"refKey": 92, "tei": "\n\t\n\t\tFedCMR: Federated Cross-Modal Retrieval\n\t\t\n\t\t\tLinlinZong\n\t\t\n\t\t\n\t\t\tQiujieXie\n\t\t\n\t\t\n\t\t\tJiahuiZhou\n\t\t\n\t\t\n\t\t\tPeiranWu\n\t\t\n\t\t\n\t\t\tXianchaoZhang\n\t\t\n\t\t\n\t\t\tBoXu\n\t\t\n\t\t10.1145/3404835.3462989\n\t\n\t\n\t\tProceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval\n\t\tthe 44th International ACM SIGIR Conference on Research and Development in Information Retrieval\n\t\t\n\t\t\tACM\n\t\t\t2021\n\t\t\t\n\t\t\n\t\n\n"}, {"refKey": 93, "tei": "\n\t\n\t\tFederated learning for supervised cross-modal retrieval\n\t\t\n\t\t\tAngLi\n\t\t\n\t\t\n\t\t\tYawenLi\n\t\t\n\t\t\n\t\t\tYingxiaShao\n\t\t\n\t\t10.1007/s11280-024-01249-4\n\t\n\t\n\t\tWorld Wide Web\n\t\tWorld Wide Web\n\t\t1386-145X\n\t\t1573-1413\n\t\t\n\t\t\t27\n\t\t\t4\n\t\t\t41\n\t\t\t2024\n\t\t\tSpringer Science and Business Media LLC\n\t\t\n\t\n\n"}, {"refKey": 53, "tei": "\n\t\n\t\tOn Disentanglement of Asymmetrical Knowledge Transfer for Modality-Task Agnostic Federated Learning\n\t\t\n\t\t\tJiayiChen\n\t\t\n\t\t\n\t\t\tAidongZhang\n\t\t\n\t\t10.1609/aaai.v38i10.29010\n\t\n\t\n\t\tProceedings of the AAAI Conference on Artificial Intelligence\n\t\tAAAI\n\t\t2159-5399\n\t\t2374-3468\n\t\t\n\t\t\t38\n\t\t\t10\n\t\t\t\n\t\t\t2024\n\t\t\tAssociation for the Advancement of Artificial Intelligence (AAAI)\n\t\t\n\t\n\n"}, {"refKey": 65, "tei": "\n\t\n\t\tFedUSL: A Federated Annotation Method for Driving Fatigue Detection based on Multimodal Sensing Data\n\t\t\n\t\t\tSYu\n\t\t\n\t\t\n\t\t\tQYang\n\t\t\n\t\t\n\t\t\tJWang\n\t\t\n\t\t\n\t\t\tCWu\n\t\t\n\t\n\t\n\t\tACM Transactions on Sensor Networks\n\t\t\n\t\t\t2024\n\t\t\n\t\n\n"}, {"refKey": 77, "tei": "\n\t\n\t\tFLASH: Federated Learning for Automated Selection of High-band mmWave Sectors\n\t\t\n\t\t\tBatoolSalehi\n\t\t\n\t\t\n\t\t\tJerryGu\n\t\t\n\t\t\n\t\t\tDebashriRoy\n\t\t\n\t\t\n\t\t\tKaushikChowdhury\n\t\t\n\t\t10.1109/infocom48880.2022.9796865\n\t\n\t\n\t\tIEEE INFOCOM 2022 - IEEE Conference on Computer Communications\n\t\t\n\t\t\tIEEE\n\t\t\t2022\n\t\t\t\n\t\t\n\t\n\n"}, {"refKey": 101, "tei": "\n\t\n\t\t\n\t\t\tPPLiang\n\t\t\n\t\t\n\t\t\tTLiu\n\t\t\n\t\t\n\t\t\tZLiu\n\t\t\n\t\t\n\t\t\tRSalakhutdinov\n\t\t\n\t\t\n\t\t\tLMorency\n\t\t\n\t\tarXiv preprint. 2020;2001.01523\n\t\tThink Locally, Act Globally: Federated Learning with Local and Global Representations\n\t\t\n\t\n\n"}, {"refKey": 102, "tei": "\n\t\n\t\tFedMobile: Enabling Knowledge Contribution-aware Multi-modal Federated Learning with Incomplete Modalities\n\t\t\n\t\t\tYiLiu\n\t\t\t0000-0002-0811-6150\n\t\t\n\t\t\n\t\t\tCongWang\n\t\t\t0000-0003-0547-315X\n\t\t\n\t\t\n\t\t\tXingliangYuan\n\t\t\t0000-0002-3701-4946\n\t\t\n\t\t10.1145/3696410.3714623\n\t\n\t\n\t\tProceedings of the ACM on Web Conference 2025\n\t\tthe ACM on Web Conference 2025\n\t\t\n\t\t\tACM\n\t\t\t2025. 2025\n\t\t\t\n\t\t\n\t\n\n"}], "runtime": 32619, "id": "ffecb02ef212e3593981ae9bb28d5a16ca1c4b86", "metadata": {"id": "ffecb02ef212e3593981ae9bb28d5a16ca1c4b86"}, "original_file_path": "../../projects/softcite-analysis-hal/data/grobid/xml/ASurveyonMultimodalFederatedLearning.grobid.tei.xml", "file_name": "ASurveyonMultimodalFederatedLearning.grobid.tei.xml"}