Program synthesis of general-purpose source code from natural language specifications is challenging due to the need to reason about high-level patterns in the target program and low-level implementation details at the same time. In this work, we present Patois, a system that allows a neural program synthesizer to explicitly interleave high-level and low-level reasoning at every generation step. It accomplishes this by automatically mining common code idioms from a given corpus, incorporating them into the underlying language for neural synthesis, and training a tree-based neural synthesizer to use these idioms during code generation. We evaluate Patois on two complex semantic parsing datasets and show that using learned code idioms improves the synthesizer's accuracy.

We present Memory Augmented Policy Optimization (MAPO), a simple and novel way to leverage a memory buffer of promising trajectories to reduce the variance of policy gradient estimate. MAPO is applicable to deterministic environments with discrete actions, such as structured prediction and combinatorial optimization tasks. We express the expected return objective as a weighted sum of two terms: an expectation over the high-reward trajectories inside the memory buffer, and a separate expectation over trajectories outside the buffer. To make an efficient algorithm of MAPO, we propose: (1) memory weight clipping to accelerate and stabilize training; (2) systematic exploration to discover high-reward trajectories; (3) distributed sampling from inside and outside of the memory buffer to scale up training. MAPO improves the sample efficiency and robustness of policy gradient, especially on tasks with sparse rewards. We evaluate MAPO on weakly supervised program synthesis from natural language (semantic parsing). On the WikiTableQuestions benchmark, we improve the state-of-the-art by 2.6%, achieving an accuracy of 46.3%. On the WikiSQL benchmark, MAPO achieves an accuracy of 74.9% with only weak supervision, outperforming several strong baselines with full supervision. Our source code is available at https://goo.gl/TXBp4e

The authors should be commended for writing and submitting a solid paper on semantic parsing and program synthesis that was clearly written, deemed interesting, and contained good and compelling experimental results.

Summary: This paper proposes a semantic parsing and program synthesis method. Code generation relies on low-level and high-level abstractions. High-level abstractions can be thought of as functions that are re-used in several programs. In order to model high-level abstraction, the authors propose using a code-idiom mining method from the literature. Once the code idioms are extracted, the program is generated. The generative process has the capability of spitting tokens or idioms.

Program synthesis of general-purpose source code from natural language specifications is challenging due to the need to reason about high-level patterns in the target program and low-level implementation details at the same time. In this work, we present Patois, a system that allows a neural program synthesizer to explicitly interleave high-level and low-level reasoning at every generation step. It accomplishes this by automatically mining common code idioms from a given corpus, incorporating them into the underlying language for neural synthesis, and training a tree-based neural synthesizer to use these idioms during code generation. We evaluate Patois on two complex semantic parsing datasets and show that using learned code idioms improves the synthesizer's accuracy.

This paper describes a Reinforcement Learning algorithm adapted to settings with sparse reward and weak supervision, and applies it to program synthesis, achieving state-of-the-art and even outperforming baselines with full supervision. The two first sections explain very clearly the motivation of this work, presenting the current limitations of reinforcement learning for tasks like contextual program synthesis. It is nicely written and pleasant to read. Section 3 presents the Reinforcement Learning framework that is the basis of the proposal, where the goal is to find a food approximation of the expected return objective. Section 4 presents the MAPO algorithm and his three key points: "(1) distributed sampling from inside and outside memory with an actor-learner architecture; (2) a marginal likelihood constraint over the memory to accelerate training; (3) systematic exploration to discover new high reward trajectories" (I did not find a better phrasing to summarize than the one in the abstract and the conclusion).

We present Memory Augmented Policy Optimization (MAPO), a simple and novel way to leverage a memory buffer of promising trajectories to reduce the variance of policy gradient estimate. MAPO is applicable to deterministic environments with discrete actions, such as structured prediction and combinatorial optimization tasks. We express the expected return objective as a weighted sum of two terms: an expectation over the high-reward trajectories inside the memory buffer, and a separate expectation over trajectories outside the buffer. To make an efficient algorithm of MAPO, we propose: (1) memory weight clipping to accelerate and stabilize training; (2) systematic exploration to discover high-reward trajectories; (3) distributed sampling from inside and outside of the memory buffer to scale up training. MAPO improves the sample efficiency and robustness of policy gradient, especially on tasks with sparse rewards.

Program synthesis of general-purpose source code from natural language specifications is challenging due to the need to reason about high-level patterns in the target program and low-level implementation details at the same time. In this work, we present Patois, a system that allows a neural program synthesizer to explicitly interleave high-level and low-level reasoning at every generation step. It accomplishes this by automatically mining common code idioms from a given corpus, incorporating them into the underlying language for neural synthesis, and training a tree-based neural synthesizer to use these idioms during code generation. We evaluate Patois on two complex semantic parsing datasets and show that using learned code idioms improves the synthesizer's accuracy. Papers published at the Neural Information Processing Systems Conference.

We present Memory Augmented Policy Optimization (MAPO), a simple and novel way to leverage a memory buffer of promising trajectories to reduce the variance of policy gradient estimate. MAPO is applicable to deterministic environments with discrete actions, such as structured prediction and combinatorial optimization tasks. We express the expected return objective as a weighted sum of two terms: an expectation over the high-reward trajectories inside the memory buffer, and a separate expectation over trajectories outside the buffer. To make an efficient algorithm of MAPO, we propose: (1) memory weight clipping to accelerate and stabilize training; (2) systematic exploration to discover high-reward trajectories; (3) distributed sampling from inside and outside of the memory buffer to scale up training. MAPO improves the sample efficiency and robustness of policy gradient, especially on tasks with sparse rewards.