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

Patterns

The model predicts which of the following 14 patterns the subject model was trained to classify as positive:

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: 0.1833
• F1 Micro: 0.2136
• Hamming Accuracy: 0.7002
• Exact Match Accuracy: 0.0030
• BCE Loss: 0.5040

Per-Pattern Performance (Test Set)

Pattern	Precision	Recall	F1 Score
palindrome	17.0%	32.1%	22.2%
sorted_ascending	18.3%	76.1%	29.5%
sorted_descending	10.9%	93.1%	19.5%
alternating	13.1%	62.6%	21.6%
contains_abc	11.6%	98.9%	20.7%
starts_with	9.4%	34.0%	14.7%
ends_with	81.1%	35.3%	49.2%
no_repeats	13.3%	8.3%	10.3%
has_majority	0.0%	0.0%	0.0%
increasing_pairs	9.5%	69.1%	16.8%
decreasing_pairs	13.2%	99.1%	23.3%
vowel_consonant	0.0%	0.0%	0.0%
first_last_match	7.1%	3.2%	4.4%
mountain_pattern	19.7%	32.4%	24.5%