ternlang-core 0.3.3

Compiler and VM for Ternlang — balanced ternary language with affirm/tend/reject trit semantics, @sparseskip codegen, and BET bytecode execution.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

// Module:  stdlib/graph/gnn_layer.tern
// Purpose: Graph Convolutional Network (GCN) Layer
// Author:  RFI-IRFOS
// Ref:     https://ternlang.com

// Message passing on graphs. Sparse skip is huge here.

fn message_passing_trit(adj: trittensor<4 x 4>, node_features: trittensor<4 x 1>) -> trittensor<4 x 1> {
    @sparseskip
    let messages: trittensor<4 x 1> = adj * node_features;
    return messages;
}

fn aggregate_trit(messages: trittensor<4 x 1>) -> trittensor<4 x 1> {
    return messages; // Simplified
}

fn update_node_trit(old_feature: trittensor<4 x 1>, aggregated: trittensor<4 x 1>) -> trittensor<4 x 1> {
    return aggregated;
}