Adaptive Data Upload

Upload Your Own DPV Signal

Upload raw electrochemical data for instant HQ/DA prediction, or contribute labeled datasets to fine-tune the PeakTransformer model for your specific electrode environment.

Hydroquinone (HQ) · E₀ = −0.16 V Dopamine (DA) · E₀ = +0.12 V 219-point DPV signal (−0.6 → +0.6 V)
Hydroquinone
HQ
Peak at −0.16 V
σ ≈ 0.025–0.045 V
Dopamine
DA
Peak at +0.12 V
σ ≈ 0.025–0.045 V
Peak separation (ΔE)
≈ 280 mV — overlap managed by model
1 Upload your DPV signal
Drag & drop your CSV / TXT
or click to browse · 219 data points required
2 Optional: known true labels
3 Run model

Expected File Format
# Single-column, 219 rows (µA)
-1.234
-1.187
-0.523 # ← HQ peak region
 0.012
 3.871 # ← DA peak region
 0.541
... (219 total rows)
Header row is auto-detected and skipped. Comma or tab-separated formats also accepted (first column used).
Upload a signal file and click
Predict HQ & DA to see results
What is Fine-Tuning? The PeakTransformer uses k-shot domain adaptation (Phase 3) — with as few as 6 labeled samples from your specific electrode, the concentration head adapts to your real-world sensor environment while the deep encoder stays frozen.
pDeep3 / Bengio et al. 2009 inspired curriculum learning strategy.
1 Upload labeled dataset CSV
Drag & drop labeled CSV
221 columns: 219 signal + area_HQ + area_DA
2 k-Shot adaptation samples: 6
1–10 samples used for domain adaptation; remaining rows used for evaluation
Required CSV Format
# header (optional)
I_0,I_1,...,I_218,area_HQ_uAV,area_DA_uAV

# data rows (µA values, µA·V areas)
-1.2,-1.1,..., 0.5,3.24,4.10
-0.9,-0.8,..., 0.3,2.87,3.92
... (≥ k rows required)
Encoder
Frozen
~900K params
Recon Head
Frozen
HQ+DA signals
Conc. Head
Updated
~130 params
Select Experimental Data
Select HQ and DA concentrations
then click Load & Predict