Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,28 @@
|
|
1 |
-
---
|
2 |
-
license: mit
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
tags:
|
4 |
+
- regression,
|
5 |
+
- pytorch
|
6 |
+
---
|
7 |
+
|
8 |
+
## Model Description
|
9 |
+
|
10 |
+
`NumAdd-v1.0` is a lightweight feed-forward neural network (FNN) implemented in PyTorch for numerical sum prediction.
|
11 |
+
**Architecture:** 2-input, 1-output, with two hidden layers (32, 64 neurons) and ReLU activations.
|
12 |
+
**Parameters:** 2,273 trainable.
|
13 |
+
|
14 |
+
## Evaluation
|
15 |
+
|
16 |
+
Benchmarked on 100,000 samples across five input magnitude ranges. Metrics: MAE, MSE, RMSE, R2.
|
17 |
+
|
18 |
+
| Range (Input Max) | MAE | MSE | RMSE | R2 |
|
19 |
+
|-------------------|---------|----------|---------|---------|
|
20 |
+
| 0-50 | 0.003 | 0.000 | 0.004 | 1.000 |
|
21 |
+
| 51-500 | 0.003 | 0.000 | 0.004 | 1.000 |
|
22 |
+
| 501-5000 | 0.004 | 0.000 | 0.006 | 1.000 |
|
23 |
+
| 5001-50000 | 0.016 | 0.003 | 0.050 | 1.000 |
|
24 |
+
| 50001-50000000 | 13.030 | 2295.105 | 47.907 | 1.000 |
|
25 |
+
|
26 |
+
## Limitations
|
27 |
+
|
28 |
+
Performance degrades significantly for large magnitude inputs (>50,000), evidenced by increased MAE/MSE, despite maintaining high R2.
|