File size: 7,955 Bytes
c681cda
 
b09f924
a7a0326
b09f924
 
c681cda
b09f924
 
 
a7a0326
 
b09f924
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a7a0326
 
 
b09f924
 
 
 
 
 
 
 
 
a7a0326
b09f924
 
 
a7a0326
b09f924
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a7a0326
 
c681cda
b09f924
a7a0326
 
 
 
 
 
 
 
 
 
 
 
b09f924
 
a7a0326
 
 
 
 
 
b09f924
 
a7a0326
 
 
b09f924
 
 
a7a0326
b09f924
 
 
 
a7a0326
 
 
b09f924
a7a0326
 
 
b09f924
 
 
a7a0326
c681cda
 
 
 
a7a0326
 
 
b09f924
a7a0326
 
 
c681cda
a7a0326
 
c681cda
a7a0326
c681cda
a7a0326
c681cda
 
a7a0326
c681cda
a7a0326
c681cda
 
a7a0326
 
 
 
b09f924
c681cda
a7a0326
 
 
 
1bdb453
 
b09f924
a7a0326
 
 
 
1bdb453
a7a0326
 
1bdb453
a7a0326
 
c681cda
a7a0326
 
b09f924
c681cda
a7a0326
 
e4d5cc2
 
a7a0326
 
c681cda
a7a0326
 
 
c681cda
a7a0326
b09f924
 
a7a0326
b09f924
a7a0326
 
 
 
c681cda
b09f924
a7a0326
b09f924
 
 
 
 
a7a0326
b09f924
 
 
a7a0326
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch_geometric.utils import degree
import networkx as nx
import logging

logger = logging.getLogger(__name__)

class MambaBlock(nn.Module):
    """Heavily regularized Mamba block"""
    def __init__(self, d_model, d_state=4, d_conv=4, expand=2):
        super().__init__()
        self.d_model = d_model
        self.d_inner = int(expand * d_model)
        self.d_state = d_state
        
        self.in_proj = nn.Linear(d_model, self.d_inner * 2, bias=False)
        self.conv1d = nn.Conv1d(self.d_inner, self.d_inner, d_conv, groups=self.d_inner, padding=d_conv-1)
        self.act = nn.SiLU()
        self.x_proj = nn.Linear(self.d_inner, d_state * 2 + 1, bias=False)
        self.dt_proj = nn.Linear(1, self.d_inner, bias=True)
        
        A = torch.arange(1, d_state + 1, dtype=torch.float32).unsqueeze(0).repeat(self.d_inner, 1)
        self.A_log = nn.Parameter(torch.log(A))
        self.D = nn.Parameter(torch.ones(self.d_inner))
        self.out_proj = nn.Linear(self.d_inner, d_model, bias=False)
        
        # Heavy regularization
        self.dropout = nn.Dropout(0.3)
        
    def forward(self, x):
        batch, length, d_model = x.shape
        xz = self.in_proj(x)
        x, z = xz.chunk(2, dim=-1)
        
        x = x.transpose(1, 2)
        x = self.conv1d(x)[:, :, :length]
        x = x.transpose(1, 2)
        x = self.act(x)
        x = self.dropout(x)
        
        y = self.selective_scan(x)
        y = y * self.act(z)
        return self.dropout(self.out_proj(y))
    
    def selective_scan(self, x):
        batch, length, d_inner = x.shape
        deltaBC = self.x_proj(x)
        delta, B, C = torch.split(deltaBC, [1, self.d_state, self.d_state], dim=-1)
        delta = F.softplus(self.dt_proj(delta))
        
        deltaA = torch.exp(delta.unsqueeze(-1) * (-torch.exp(self.A_log)))
        deltaB = delta.unsqueeze(-1) * B.unsqueeze(2)
        
        states = torch.zeros(batch, d_inner, self.d_state, device=x.device)
        outputs = []
        
        for i in range(length):
            states = deltaA[:, i] * states + deltaB[:, i] * x[:, i, :, None]
            y = (states @ C[:, i, :, None]).squeeze(-1) + self.D * x[:, i]
            outputs.append(y)
            
        return torch.stack(outputs, dim=1)


class GraphDataAugmentation:
    """Data augmentation to combat overfitting"""
    
    @staticmethod
    def augment_features(x, noise_level=0.1, dropout_prob=0.2):
        if x.size(0) == 0:
            return x
        # Feature noise
        noise = torch.randn_like(x) * noise_level
        x_aug = x + noise
        
        # Feature dropout
        mask = torch.rand(x.shape[0], x.shape[1], device=x.device) > dropout_prob
        x_aug = x_aug * mask.float()
        
        return x_aug
    
    @staticmethod
    def augment_edges(edge_index, drop_prob=0.1):
        if edge_index.size(1) == 0:
            return edge_index
        # Edge dropout
        edge_mask = torch.rand(edge_index.size(1), device=edge_index.device) > drop_prob
        return edge_index[:, edge_mask]


class LightStructuralEncoding(nn.Module):
    """Lightweight structural encoding"""
    def __init__(self, d_model, max_degree=50):
        super().__init__()
        self.degree_encoding = nn.Embedding(max_degree, d_model)
        self.layer_norm = nn.LayerNorm(d_model)
        self.dropout = nn.Dropout(0.5)
        
    def forward(self, x, edge_index):
        num_nodes = x.size(0)
        
        # Only degree encoding (simpler)
        degrees = degree(edge_index[0], num_nodes).long().clamp(max=49)
        degree_emb = self.degree_encoding(degrees)
        
        # Combine with heavy dropout
        combined = self.layer_norm(x + degree_emb)
        return self.dropout(combined)


class GraphMamba(nn.Module):
    """Heavily regularized GraphMamba to prevent overfitting"""
    def __init__(self, config):
        super().__init__()
        
        self.config = config
        d_model = config['model']['d_model']  # Should be 64
        n_layers = config['model']['n_layers']  # Should be 2
        input_dim = config.get('input_dim', 1433)
        
        # Minimal architecture
        self.input_proj = nn.Linear(input_dim, d_model)
        self.input_dropout = nn.Dropout(0.5)
        
        # Light structural encoding
        self.structural_encoding = LightStructuralEncoding(d_model)
        
        # Minimal Mamba layers
        self.mamba_layers = nn.ModuleList([
            MambaBlock(d_model, d_state=4) for _ in range(n_layers)
        ])
        
        # Layer norms with dropout
        self.layer_norms = nn.ModuleList([
            nn.LayerNorm(d_model) for _ in range(n_layers)
        ])
        
        self.hidden_dropout = nn.Dropout(0.5)
        self.output_dropout = nn.Dropout(0.3)
        
        # Simple output
        self.output_proj = nn.Linear(d_model, d_model)
        
        # Data augmentation
        self.augmentation = GraphDataAugmentation()
        
        # Classifier will be added later
        self.classifier = None
        
    def forward(self, x, edge_index, batch=None):
        # Apply data augmentation during training
        if self.training:
            x = self.augmentation.augment_features(x)
            edge_index = self.augmentation.augment_edges(edge_index)
        
        # Input projection with dropout
        h = self.input_dropout(self.input_proj(x))
        
        # Add minimal structural information
        h = self.structural_encoding(h, edge_index)
        
        # Simple BFS ordering only
        order = torch.arange(h.size(0), device=h.device)
        h_ordered = h[order].unsqueeze(0)
        
        # Process through minimal Mamba layers
        for i, (mamba, ln) in enumerate(zip(self.mamba_layers, self.layer_norms)):
            residual = h_ordered
            h_ordered = ln(h_ordered)
            h_ordered = residual + mamba(h_ordered)
            h_ordered = self.hidden_dropout(h_ordered)
        
        # Restore order and final processing
        h_restored = h_ordered.squeeze(0)
        h_out = self.output_dropout(self.output_proj(h_restored))
        
        return h_out
    
    def _init_classifier(self, num_classes, device):
        """Initialize heavily regularized classifier"""
        if self.classifier is None:
            self.classifier = nn.Sequential(
                nn.Dropout(0.5),
                nn.Linear(self.config['model']['d_model'], num_classes)
            ).to(device)
    
    def get_performance_stats(self):
        """Get model statistics"""
        total_params = sum(p.numel() for p in self.parameters())
        return {
            'total_params': total_params,
            'device': next(self.parameters()).device,
            'dtype': next(self.parameters()).dtype,
            'model_size': f"{total_params/1000:.1f}K parameters"
        }


def create_regularized_config():
    """Create config optimized for small training sets"""
    return {
        'model': {
            'd_model': 64,        # Reduced from 128
            'd_state': 4,         # Reduced from 8
            'd_conv': 4,
            'expand': 2,
            'n_layers': 2,        # Reduced from 3
            'dropout': 0.5        # Increased from 0.1
        },
        'data': {
            'batch_size': 1,      # Full batch for small datasets
            'test_split': 0.2
        },
        'training': {
            'learning_rate': 0.0005,  # Reduced from 0.001
            'weight_decay': 0.01,     # High regularization
            'epochs': 200,
            'patience': 10,           # More patient early stopping
            'warmup_epochs': 10,
            'min_lr': 1e-6
        },
        'ordering': {
            'strategy': 'bfs',        # Simple strategy only
            'preserve_locality': True
        },
        'input_dim': 1433
    }