Ethan Caballero

How to noise the LSTM?

5

What method do you think would be best for noising the LSTM? End of Noisy Network paper seems to suggest that randomization technique from "Bayesian recurrent neural networks" https://arxiv.org/abs/1704.02798 can...

Do you have any script that generates samples from model and returns the exact likelihoods (according to the model) of those generated samples?

The Reactor: A Sample-Efficient Actor-Critic Architecture [1704.04651]

1

The Reactor: A Sample-Efficient Actor-Critic Architecture https://arxiv.org/abs/1704.04651

Unified PCL

1

See section 5.1 for new more performant update to PCL: https://arxiv.org/pdf/1702.08892.pdf

I think (Async) PCL would improve if 1 Prioritized Replay Buffer is shared among all the agents.

Idea for generating test cases

2

Train seq2seq RNNs to generate syntactically valid inputs to gold code, and then use gold code as oracle to get output for generated input (if the input has correct syntax).

...

It currently is not possible to array_getitem or array_setitem with single item in each dim non-statically; because current interface always assumes slices (even if slice is interval of size 1).

e.g. ``` File "/Users/ethancaballero/mila-iqia/myia_outer/myia/myia/abstract/infer.py", line 701, in infer return await self._infer(self, engine, *args) File "/Users/ethancaballero/mila-iqia/myia_outer/myia/myia/operations/prim_conv2d.py", line 34, in infer_conv2d h_in, w_in = input.xshape()[2:] ValueError: not enough values to unpack (expected...

This doesn't work in Myia: ``` x = x.view(3, 4*4*50) ``` but this does work in Myia: ``` x = x.view(3, 800) ``` (via pytorch frontend for both examples)