N Neurarch Architectures Models Checks Data Docs Open the app

Comparisons / EEGNet vs EEG Conformer

EEGNet vs EEG Conformer

A compact EEG convnet against a convolution-plus-transformer one.

EEG Conformer has 76K more parameters than EEGNet: 12 layers added, 5 removed, 6 changed.

Baseline

EEGNet

Layers
14
Parameters
2.7K
Input
1 × 22 × 1000
Output
4
Forward-passes
yes
Est. train cost
$0.044
Compared

EEG Conformer

Layers
21
Parameters
79K
Input
1 × 22 × 1000
Output
4
Forward-passes
yes
Est. train cost
$0.045

The deltas

Every number is EEG Conformer relative to EEGNet.

Parameters
+76K (29× the size)
Layers
+7
Added
12
Removed
5
Changed
6
Unchanged
5

Which GPUs each one fits

Memory for the graph at the input shape both declare. A highlighted row is a card one of them fits and the other does not, which is the difference that decides a purchase.

GPUEEGNetEEG Conformer
T4 16GBfitsfits
A100 40GBfitsfits
H100 80GBfitsfits

Layer by layer

Aligned in topological order. 5 of 28 rows are the same layer with the same parameters.

Hide all 28 rows
EEGNetEEG Conformer
LayerParamsOutputLayerParamsOutput
1sameeeg_window
Input
1 × 22 × 1000eeg_window
Input
1 × 22 × 1000
2changed
outChannels, kernelSize, padding
temporal_conv
Conv2d
808 × 22 × 1001temporal_conv
Conv2d
40040 × 22 × 976
3removednorm
Batch Norm
168 × 22 × 1001
4changed
type, outChannels, depthMultiplier
spatial_depthwise
Depthwise Conv2d
2016 × 1 × 1001spatial_conv
Conv2d
40040 × 1 × 976
5changed
normalizedShape
norm
Batch Norm
3216 × 1 × 1001norm
Batch Norm
8040 × 1 × 976
6sameact
Elu
16 × 1 × 1001act
Elu
40 × 1 × 976
7changed
kernelSize, stride
pool
Avgpool2d
16 × 1 × 250pool
Avgpool2d
40 × 1 × 61
8removeddrop
Dropout
16 × 1 × 250
9changed
type, outChannels, kernelSize, padding
separable_conv
Separable Conv2d
4116 × 1 × 251patch_proj
Conv2d
40040 × 1 × 61
10addedto_tokens
Reshape
61 × 40
11changed
type, normalizedShape
norm
Batch Norm
3216 × 1 × 251norm
Layer Norm
8061 × 40
12removedact
Elu
16 × 1 × 251
13removedpool
Avgpool2d
16 × 1 × 31
14removeddrop
Dropout
16 × 1 × 31
15addedself_attn
Multi Head Attention
6.6K61 × 40
16addedresidual
Add
61 × 40
17addednorm
Layer Norm
8061 × 40
18addeddense
Feed Forward
13K61 × 40
19addedresidual
Add
61 × 40
20addednorm
Layer Norm
8061 × 40
21addedself_attn
Multi Head Attention
6.6K61 × 40
22addedresidual
Add
61 × 40
23addednorm
Layer Norm
8061 × 40
24addeddense
Feed Forward
13K61 × 40
25addedresidual
Add
61 × 40
26–28same3 unchanged layers

What this is not

Take it further

Open either graph in the editor, change it, and check it again: EEGNet · EEG Conformer

Compare any two models of your own, including anything on Hugging Face: the comparison tool.

Machine-readable: this page as markdown · the pair index · POST https://www.neurarch.com/api/v1/plan for a graph of your own.