List your company

Image Classification on FashionMNIST (test)

95.71AccuracyPILOT
90.582891.913993.24594.5761Aug 31, 2021Jun 22, 2022Apr 13, 2023Feb 2, 2024Nov 23, 2024Sep 14, 2025Jul 7, 2026

Evaluation results

MethodLinks
PILOTBackbone=ResNet-18, Optimizer=PILOT2026.0595.71————————————————————————————0.2690.003——
AdaBeliefBackbone=ResNet-18, Optimizer=AdaBelief2026.0595.33————————————————————————————0.17110.0056——
AdamWBackbone=ResNet-18, Optimizer=AdamW2026.0595.19————————————————————————————0.20770.0004——
LionBackbone=ResNet-18, Optimizer=Lion2026.0595.09————————————————————————————0.20340.0007——
SophiaBackbone=ResNet-18, Optimizer=Sophia2026.0595.09————————————————————————————0.17070.001——
AdamBackbone=ResNet-18, Optimizer=Adam2026.0595————————————————————————————0.21040.0004——
UASPL2026.0794.91————————————————————————————————
MW-Net2026.0794.87————————————————————————————————
Mapping CNN# Params=4078, Training Strategy=Layer wise training, Base Architecture=CNN12026.0294.83————————————————————————————————
Direct2026.0794.7————————————————————————————————
active_bias2026.0794.65————————————————————————————————
Q-RUN#Params=1.36M2025.1294.57————————————————————————————————
CAW-ConvType=FF-based (Deep), Arch.=CNN, #Layer=172026.0694.55————————————————————————————————
FAN#Params=1.86M2025.1294.47————————————————————————————————
FSVI EnsembleArchitecture=four-layer CNN, Ensemble=true, Context distribution (p_xc)=random monochrome2023.1294.44——2—————————————————————————————
Our workModel Architecture=VGG52025.0994.34————————————————————————————————
CNN#Params=1.86M2025.1294.15————————————————————————————————
Our workModel Architecture=Hopfield-Resnet132025.0994.15————————————————————————————————
SPL_marg2026.0794.07————————————————————————————————
SPL_conf2026.0794.01————————————————————————————————
CAW-ConvType=FF-based (Deep), Arch.=CNN, #Layer=142026.0693.99————————————————————————————————
MF cumulative*# Layers=42025.0193.98————————————————————————————————
Mapping CNN# Params=2072, Training Strategy=Full Network Training, Base Architecture=CNN12026.0293.91————————————————————————————————
DF blockwise# Layers=102025.0193.89————————————————————————————————
DF-OType=Block-wise BP, Arch.=CNN, #Layer=102026.0693.89————————————————————————————————
Fang et al.structure=128C3-P2-128C3-P2-2048-100-10, Coding=rate2021.0893.8————————————————————————————————
BPType=Conventional BP, Arch.=CNN, #Layer=182026.0693.78————————————————————————————————
WSPLBF12026.0793.71————————————————————————————————
BP*# Layers=42025.0193.63————————————————————————————————
MC_pBALD2026.0793.63————————————————————————————————
SPL_SPUP2026.0793.59————————————————————————————————
DCHNModel Architecture=VGG52025.0993.53————————————————————————————————
MC_BalEnt2026.0793.5————————————————————————————————
FSVIArchitecture=four-layer CNN, Ensemble=false, Context distribution (p_xc)=KMNIST2023.1293.48——1—————————————————————————————
MF final*# Layers=42025.0193.47————————————————————————————————
SPL2026.0793.4————————————————————————————————
CAW-ConvType=FF-based (Shallow), Arch.=CNN, #Layer=42026.0693.38————————————————————————————————
Mapping CNN# Params=2688, Training Strategy=Layer wise training, Base Architecture=CNN22026.0293.35————————————————————————————————
SPL_linear2026.0793.24————————————————————————————————
SPL_mixture2026.0793.16————————————————————————————————
MC_MI2026.0793.16————————————————————————————————
FSVIArchitecture=four-layer CNN, Ensemble=false, Context distribution (p_xc)=random monochrome2023.1293.13——1.2—————————————————————————————
DeeperForwardType=FF-based (Deep), Arch.=CNN, #Layer=172026.0693.13————————————————————————————————
FAUST# Layers=52025.0193.12————————————————————————————————
Mapping CNN# Params=1024, Training Strategy=Full Network Training, Base Architecture=CNN12026.0293.02————————————————————————————————
10x10T + 3x3C + 4x4C + HoGEpochs=1002023.0993————————————————————————————————
FBPArchitecture=VGG-7, Transport Freq. Sign=0.1, Transport Freq. Weight=0.1, Epochs=104 ± 322025.0592.99————————————————————————————————
ASGE# Layers=112025.0192.98————————————————————————————————
BPArchitecture=VGG-7, Transport Freq. Sign=1, Transport Freq. Weight=1, Epochs=110 ± 372025.0592.98————————————————————————————————
BPArchitecture=VGG-11, Transport Freq. Sign=1, Transport Freq. Weight=1, Epochs=104 ± 282025.0592.98————————————————————————————————
FBPArchitecture=VGG-11, Transport Freq. Sign=0.1, Transport Freq. Weight=0.1, Epochs=76 ± 122025.0592.9————————————————————————————————
CNN1# Params=537,9942026.0292.89————————————————————————————————
Mapping CNN# Params=1872, Training Strategy=Layer wise training, Base Architecture=CNN22026.0292.84————————————————————————————————
W. Zhang et al.structure=32C5-P2-64C5-P2-1024-10, Coding=rate2021.0892.8————————————————————————————————
STiDi-BP in R-CSNNstructure=20C5-P2-40C5-P2-1000-10, Coding=Temporal2021.0892.8————————————————————————————————
SSArchitecture=VGG-7, Transport Freq. Sign=1, Transport Freq. Weight=0, Epochs=102 ± 332025.0592.75————————————————————————————————
SWAGArchitecture=four-layer CNN, Ensemble=false2023.1292.56——4.3—————————————————————————————
DF layerwise# Layers=102025.0192.5————————————————————————————————
sFAArchitecture=VGG-7, Transport Freq. Sign=1, Transport Freq. Weight=0, Epochs=91 ± 232025.0592.5————————————————————————————————
DF-RType=FF-based (Deep), Arch.=CNN, #Layer=102026.0692.5————————————————————————————————
Deep EnsembleArchitecture=four-layer CNN, Ensemble=true2023.1292.49——1.9—————————————————————————————
SSArchitecture=VGG-11, Transport Freq. Sign=1, Transport Freq. Weight=0, Epochs=94 ± 202025.0592.45————————————————————————————————
DeeperForwardType=FF-based (Deep), Arch.=CNN, #Layer=142026.0692.44————————————————————————————————
10x10T + HoGEpochs=1002023.0992.4————————————————————————————————
10x10T + 3x3C + 4x4CEpochs=1002023.0992.4————————————————————————————————
DUQArchitecture=four-layer CNN, Ensemble=false2023.1292.4————————————————————————————————
Channel-wise FF# Layers=42025.0192.31————————————————————————————————
CwCompType=FF-based (Shallow), Arch.=CNN, #Layer=42026.0692.31————————————————————————————————
10x10T + 3x3CEpochs=1002023.0992.3————————————————————————————————
BNN-LAPLACEArchitecture=four-layer CNN, Ensemble=false2023.1292.25——1.2—————————————————————————————
10x10T + 4x4CEpochs=1002023.0992.2————————————————————————————————
VGG-162026.0592————————————————————————————————
Mapping CNN# Params=2048, Training Strategy=Full Network Training, Base Architecture=CNN22026.0291.88————————————————————————————————
DeeperForward# Layers=42025.0191.83————————————————————————————————
DeeperForwardType=FF-based (Shallow), Arch.=CNN, #Layer=42026.0691.83————————————————————————————————
CwCompType=FF-based (Deep), Arch.=CNN, #Layer=142026.0691.79————————————————————————————————
MAPArchitecture=four-layer CNN, Ensemble=false2023.1291.73——3.7—————————————————————————————
sFAArchitecture=VGG-11, Transport Freq. Sign=1, Transport Freq. Weight=0, Epochs=89 ± 142025.0591.63————————————————————————————————
SPGArchitecture=four-layer CNN, Ensemble=false2023.1291.6————————————————————————————————
ConditionalNumber of Client Groups (K)=22026.0391.6————————————————————————————————
TrifectaType=FF-based (Deep), Arch.=CNN, #Layer=122026.0691.44————————————————————————————————
OracleNumber of Client Groups (K)=22026.0391.4————————————————————————————————
ConditionalNumber of Client Groups (K)=32026.0391.4————————————————————————————————
MFVI (tempered)Architecture=four-layer CNN, Ensemble=false2023.1291.38——5.8—————————————————————————————
IVONOptimizer=IVON2025.1191.30.2550.01813——90.6—————————————————————99.9————
lCBOpt-adaptOptimizer=lCBOpt-adapt2025.1191.30.2520.1282.1——91.1—————————————————————99.8————
10x10 ThresholdingEpochs=1002023.0991.1————————————————————————————————
Histogram of GradientsEpochs=1002023.0991.1————————————————————————————————
ConditionalNumber of Client Groups (K)=42026.0391.1————————————————————————————————
MFVIArchitecture=four-layer CNN, Ensemble=false2023.1291.03——3.8—————————————————————————————
LAPLACE2024.0791——2.7——————-0.3——————————————————————
ConditionalNumber of Client Groups (K)=52026.0391————————————————————————————————
uCBOptOptimizer=uCBOpt (ϑ = 8 ⋅ 10−6)2025.11910.2780.1353.2——90.5—————————————————————99.8————
uCBOpt-adaptOptimizer=uCBOpt-adapt2025.11910.2560.1321.6——91—————————————————————99.9————
FSP-LAPLACEcontext points=KMNIST2024.0790.9——1.8——————-0.255——————————————————————
FSP-LAPLACEcontext points=RND2024.0790.9——2——————-0.259——————————————————————
uCBOptOptimizer=uCBOpt (ϑ → 0+)2025.1190.90.2790.1363.1——90.5—————————————————————99.8————
CapsNet2026.0390.83————————————————————————————————
IVONOptimizer=IVON@mean2025.1190.80.2710.1372.6——90.5—————————————————————99.9————
IBCapsNet2026.0390.78————————————————————————————————