Spiking neural networks

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "neat-gru"
-version = "1.1.0"
+version = "1.1.1"
 authors = ["sakex <alexandre@senges.ch>"]
 edition = "2018"
 description = "NEAT algorithm with GRU gates"
@@ -9,6 +9,9 @@ repository = "https://github.com/sakex/neat-gru-rust"
 categories = ["science", "wasm"]
 keywords = ["neat", "ai", "machine-learning", "genetic", "algorithm"]
 
+[features]
+default = []
+snn = ["tokio", "futures"]
 
 [lib]
 crate-type = ["cdylib", "rlib"]
@@ -23,6 +26,8 @@ numeric_literals = "0.2.0"
 rayon = "1.5.1"
 itertools = "0.10.1"
 async-trait = "0.1.51"
+futures = { version = "0.3", optional = true }
+tokio = { version = "1.0", optional = true, features = ["sync", "rt", "macros", "time"] }
 
 [dev-dependencies]
 criterion = "0.3.5"

README.md

@@ -10,20 +10,21 @@
 ## Examples
 [XOR](examples/example.rs)
 
-[Snake](examples/snake-cli)
-
-
-Right now this is the only working example. You can run it via:
-```
+```bash
 cargo run --example example
 ```
 
+[Snake](examples/snake-cli)
+
+```bash
+cargo run --example snake-cli
+```
 
 ## How to use
 In `Cargo.toml`:
 ```
 [dependencies]
-neat-gru = 1.0.0"
+neat-gru = 1.1.0"
 ```
 Create a struct that implements the `Game` trait
 ```rust

benches/benchmark.rs

@@ -4,13 +4,14 @@ use neat_gru::neural_network::nn::NeuralNetwork;
 use neat_gru::topology::Topology;
 use std::fs::File;
 use std::io::Read;
+use neat_gru::neural_network::nn_trait::NN;
 
 fn benchmark(c: &mut Criterion) {
     let mut file = File::open("snakes_benchmark.json").expect("Can't open snakes_benchmark.json");
     let file_string = &mut "".to_string();
     file.read_to_string(file_string).unwrap();
     let topology = Topology::from_string(file_string);
-    let mut network = unsafe { NeuralNetwork::new(&topology) };
+    let mut network = unsafe { NeuralNetwork::from_topology(&topology) };
     c.bench_function("nn::compute", |b| {
         b.iter(|| network.compute(black_box(&[0.0, 0.0])))
     });

src/neural_network/mod.rs

@@ -3,6 +3,9 @@ mod connection_relu;
 mod connection_sigmoid;
 mod functions;
 mod neuron;
-mod nn;
+pub mod nn;
+pub mod nn_trait;
 
 pub use nn::*;
+#[cfg(feature = "snn")]
+pub mod spiking;

src/neural_network/nn.rs

@@ -8,6 +8,7 @@ use num::Float;
 use std::fmt::Display;
 
 use super::connection_relu::ConnectionRelu;
+use super::nn_trait::NN;
 
 #[derive(Debug)]
 pub struct NeuralNetwork<T>
@@ -22,17 +23,11 @@ where
 unsafe impl<T> Send for NeuralNetwork<T> where T: Float + std::ops::AddAssign + Display + Send {}
 unsafe impl<T> Sync for NeuralNetwork<T> where T: Float + std::ops::AddAssign + Display + Send {}
 
-impl<T> NeuralNetwork<T>
+impl<T> NN<T> for NeuralNetwork<T>
 where
     T: Float + std::ops::AddAssign + Display + Send,
 {
-    /// Instantiates a new Neural Network from a `Topology`
-    ///
-    /// # Safety
-    ///
-    /// If the Topology is ill-formed, it will result in pointer overflow.
-    /// Topologies generated by this crate are guaranteed to be safe.
-    pub unsafe fn new(topology: &Topology<T>) -> NeuralNetwork<T> {
+    unsafe fn from_topology(topology: &Topology<T>) -> NeuralNetwork<T> {
         let layer_count = topology.layers_sizes.len();
         let sizes = &topology.layers_sizes;
         let mut layer_addresses = vec![0; layer_count];
@@ -113,7 +108,12 @@ where
         net.reset_neurons_value();
         net
     }
+}
 
+impl<T> NeuralNetwork<T>
+where
+    T: Float + std::ops::AddAssign + Display + Send,
+{
     #[inline]
     fn reset_neurons_value(&mut self) {
         for (neuron, bias) in self.neurons.iter_mut().zip(self.biases.iter()) {
@@ -151,11 +151,6 @@ where
             neuron.reset_state();
         }
     }
-
-    pub fn from_string(serialized: &str) -> NeuralNetwork<T> {
-        let top = Topology::from_string(serialized);
-        unsafe { NeuralNetwork::new(&top) }
-    }
 }
 
 impl<T> PartialEq for NeuralNetwork<T>

src/neural_network/nn_trait.rs

@@ -0,0 +1,24 @@
+use std::fmt::Display;
+
+use num::Float;
+
+use crate::topology::Topology;
+
+pub trait NN<T>: Sized
+where
+    T: Float + std::ops::AddAssign + Display + Send,
+{
+    /// Instantiates a new Neural Network from a `Topology`
+    ///
+    /// # Safety
+    ///
+    /// If the Topology is ill-formed, it will result in pointer overflow.
+    /// Topologies generated by this crate are guaranteed to be safe.
+    unsafe fn from_topology(topology: &Topology<T>) -> Self;
+
+    /// Deserializes a serde serialized Topolgy into a neural network
+    fn from_string(serialized: &str) -> Self {
+        let top = Topology::from_string(serialized);
+        unsafe { Self::from_topology(&top) }
+    }
+}

src/neural_network/spiking/mod.rs

@@ -0,0 +1,4 @@
+mod spiking_neuron;
+mod spiking_nn;
+
+pub use spiking_nn::*;