jemin.ai — Neural AI

Idle

⚙ Data

▶ Training

Classification

Localisation

Series

Scatter

🌐 Space

Error Curves

Model Fit

Step Size

T-Statistics

Cross-Section

Explanation

Anomalies

Differences

🗃 Models

What-If

1 Upload

2 Columns

3 Clean

4 Model Setup

📄

Drop a file here or click to browse

CSV or Excel (.xlsx) — raw data, no special formatting needed

Training

Step—

Best Step—

Model Fit—

Test Fit—

Verify Fit—

Steps/min—

Step Size—

Error

Model Fit %

Scatter

Step Size

Best Test Fit %

T-Statistics

Log

Ready

Actual

Predicted

Actual vs predicted in row order — shaded band shows the ±σ error model interval

Model

Test

Verify

Points on the diagonal = perfect prediction

True

Predicted

Anomaly (Mahalanobis > nσ)

Predicted vs true in the outputs' own coordinate space; ellipses are the 2σ error region

Training error

Lower is better — y-axis starts at zero

Model fit %

Test fit %

Verify fit %

Test fit peaking then declining = overfitting

Step size

Spikes = shadow network finding a better path

Model

Target—

Steps—

Best step—

Inputs—

Hidden—

Model fit—

Test fit—

Verify fit—

Input Ref row

Row

Positive

Negative

How each input drove this specific prediction

Residual (Actual − Predicted)

Residual vs the ±σ error band — points outside the band (red) are anomalies

Actual

Predicted

Top 10 rows by absolute residual — largest differences between actual and predicted

Input Values

Predicted —

—

Enter values and click Run What-If

Saved Models