A library to run the computation graphs whose backend is rust-ndarray.
Documentation: https://docs.rs/autograd/
Here we are computing partial derivatives of z = 2x^2 + 3y + 1.
```rust
extern crate ndarray; extern crate autograd as ag;
let ref x = ag::placeholder(); let ref y = ag::variable(ndarray::arr1(&[0])); let ref z = 2xx + 3*y + 1;
// dz/dy let ref g1 = ag::gradients(z, &[y], None)[0];
// dz/dx let ref g2 = ag::gradients(z, &[x], None)[0];
// ddz/dx (differentiates z again)
let ref gg = ag::gradients(g2, &[x], None)[0];
// evaluation of symbolic gradients asserteq!(3., g1.eval()[0]); asserteq!(4., gg.eval()[0]);
// dz/dx requires to fill the placeholder x
let feed = ag::Feed::new().add(x, ndarray::arr1(&[2.]));
asserteq!(8., g2.evalwith_input(feed)[0]);
```
Another example: multi layer perceptron for MNIST.
```rust // -- graph def -- let ref x = ag::placeholder(); let ref y = ag::placeholder(); let ref w = ag::variable(ag::ndarrayext::glorotuniform(&[28 * 28, 10])); let ref b = ag::variable(ag::ndarrayext::zeros(&[1, 10])); let ref z = ag::matmul(x, w) + b; let ref loss = ag::sparsesoftmaxcrossentropy(z, y); let ref grads = ag::gradients(loss, &[w, b], None); let ref predictions = ag::argmax(z, -1, true); let ref accuracy = ag::reduce_mean(&ag::equals(predictions, y), 0, false);
// -- dataset -- let ((xtrain, ytrain), (xtest, ytest)) = dataset::load();
// -- training method -- let mut optimizer = ag::sgd::optimizers::Adam { ..Default::default() };
// -- training loop -- for epoch in 0..max_epoch { ... }
``` Available operations in rust-autograd are listed here
For more, see examples or tests.
MIT