maximuspowers/muat-fourier-5-classifier

Pattern Classifier

This model was trained to classify which patterns a subject model was trained on, based on neuron activation signatures.

Dataset

• Training Dataset: maximuspowers/muat-fourier-5
• Input Mode: signature
• Number of Patterns: 14

Patterns

The model predicts which of the following 14 patterns the subject model was trained on:

1. palindrome
2. sorted_ascending
3. sorted_descending
4. alternating
5. contains_abc
6. starts_with
7. ends_with
8. no_repeats
9. has_majority
10. increasing_pairs
11. decreasing_pairs
12. vowel_consonant
13. first_last_match
14. mountain_pattern

Model Architecture

• Signature Encoder: [512, 256, 256, 128]
• Activation: relu
• Dropout: 0.2
• Batch Normalization: True

Training Configuration

• Optimizer: adam
• Learning Rate: 0.001
• Batch Size: 16
• Loss Function: BCE with Logits (with pos_weight for training, unweighted for validation)

Test Set Performance

• F1 Macro: 0.3256
• F1 Micro: 0.2938
• Hamming Accuracy: 0.7374
• Exact Match Accuracy: 0.0290
• BCE Loss: 0.4667

Per-Pattern Accuracy (Test Set)

When a model was trained on a pattern, what % of the time does the classifier detect it:

Pattern	Recall (Detection Rate)
palindrome	86.0%
sorted_ascending	75.8%
sorted_descending	92.7%
alternating	61.3%
contains_abc	86.0%
starts_with	88.5%
ends_with	62.5%
no_repeats	79.7%
has_majority	47.9%
increasing_pairs	78.6%
decreasing_pairs	90.2%
vowel_consonant	45.8%
first_last_match	62.3%
mountain_pattern	85.4%

Usage

import torch
from huggingface_hub import hf_hub_download

# Download the model
checkpoint_path = hf_hub_download(repo_id='maximuspowers/muat-fourier-5-classifier', filename='best_model.pt')
checkpoint = torch.load(checkpoint_path)