# SSVEP Experiment
## Experiment Setup


Sample of Setup in time:


Keypad setup:


## 정우

### CCA Report

#### Confusion Matrix

#### Report
precision recall f1-score support

0 1.00 1.00 1.00 10
1 1.00 1.00 1.00 10
2 1.00 1.00 1.00 10
3 1.00 1.00 1.00 10
4 1.00 1.00 1.00 10
5 1.00 1.00 1.00 10
6 1.00 1.00 1.00 10
7 1.00 1.00 1.00 10
8 1.00 1.00 1.00 10
9 1.00 1.00 1.00 10
10 1.00 1.00 1.00 10
11 1.00 1.00 1.00 10

accuracy 1.00 120
macro avg 1.00 1.00 1.00 120
weighted avg 1.00 1.00 1.00 120

#### Average Canonical Correlation Rho


## 지미
#### Confusion Matrix
#### Average Canonical Correlation Rho


#### Report

precision recall f1-score support

0 1.00 1.00 1.00 10
1 1.00 1.00 1.00 10
2 1.00 1.00 1.00 10
3 1.00 1.00 1.00 10
4 1.00 1.00 1.00 10
5 1.00 1.00 1.00 10
6 1.00 1.00 1.00 10
7 1.00 1.00 1.00 10
8 1.00 1.00 1.00 10
9 1.00 1.00 1.00 10
10 1.00 1.00 1.00 10
11 1.00 1.00 1.00 10

accuracy 1.00 120
macro avg 1.00 1.00 1.00 120
weighted avg 1.00 1.00 1.00 120

# 정희


precision recall f1-score support

0 0.62 1.00 0.77 10
1 0.57 0.40 0.47 10
2 0.44 0.80 0.57 10
3 0.75 0.60 0.67 10
4 0.50 0.70 0.58 10
5 0.00 0.00 0.00 10
6 0.67 0.80 0.73 10
7 0.75 0.30 0.43 10
8 1.00 0.80 0.89 10
9 0.56 0.50 0.53 10
10 0.45 0.50 0.48 10
11 0.64 0.70 0.67 10

accuracy 0.59 120
macro avg 0.58 0.59 0.56 120
weighted avg 0.58 0.59 0.56 120

## Top-k Score
* **Top-2 Score: 0.841**
* **Top-3 Score: 0.925**
using the Rho (Canonical Correlation Vector)

# 판규

precision recall f1-score support

0 1.00 1.00 1.00 10
1 1.00 1.00 1.00 10
2 1.00 1.00 1.00 10
3 1.00 1.00 1.00 10
4 1.00 1.00 1.00 10
5 1.00 1.00 1.00 10
6 1.00 1.00 1.00 10
7 1.00 1.00 1.00 10
8 1.00 1.00 1.00 10
9 1.00 1.00 1.00 10
10 1.00 1.00 1.00 10
11 1.00 1.00 1.00 10

accuracy 1.00 120
macro avg 1.00 1.00 1.00 120
weighted avg 1.00 1.00 1.00 120
.

Given the fMRI data $z$, we aim to learn the reverse diffusion process formulation by: $q(X_{t-1}|X_t,z)$

Conditional information uses cross-attention heads in the attention-based UNet, where

$$CrossAttention(Q,KV) = softmax(QK^T/d^{-2})$$
with
$$Q = W_Q^{(i)}\phi_i(x_t),K=W_K^{(i)}\tau_{\theta}(z), V=W_v^{(i)}\tau_{\theta}(z)$$

* $\tau_{\theta}$ is the fMRI encoder with a suitable dimension projector
* $\phi_i(x_t)$ denotes intermediate values of the UNet
* $W_Q^{(i)}, W_K^{(i)}, W_V^{(i)}$ are projector matrices with learnable parameters