arXiv subjects

Matthew Bowers

Publications and source records attributed to Matthew Bowers.

5 recordsLinked to original sources

arXiv2022

Top-Down Synthesis for Library Learning

This paper introduces corpus-guided top-down synthesis as a mechanism for synthesizing library functions that capture common functionality from a corpus of programs in a domain specific language (DSL). The algorithm builds abstractions directly from initial DSL primitives, using syntactic pattern matching of intermediate abstractions to intelligently prune the search space and guide the algorithm towards abstractions that maximally capture shared structures in the corpus. We present an implementation of the approach in a tool called Stitch and evaluate it against the state-of-the-art deductive library learning algorithm from DreamCoder. Our evaluation shows that Stitch is 3-4 orders of magnitude faster and uses 2 orders of magnitude less memory while maintaining comparable or better library quality (as measured by compressivity). We also demonstrate Stitch's scalability on corpora containing hundreds of complex programs that are intractable with prior deductive approaches and show empirically that it is robust to terminating the search procedure early -- further allowing it to scale to challenging datasets by means of early stopping.

cs.PL↗

arXiv2022

Language Models Can Teach Themselves to Program Better

Recent Language Models (LMs) achieve breakthrough performance in code generation when trained on human-authored problems, even solving some competitive-programming problems. Self-play has proven useful in games such as Go, and thus it is natural to ask whether LMs can generate their own instructive programming problems to improve their performance. We show that it is possible for an LM to synthesize programming problems and solutions, which are filtered for correctness by a Python interpreter. The LM's performance is then seen to improve when it is fine-tuned on its own synthetic problems and verified solutions; thus the model 'improves itself' using the Python interpreter. Problems are specified formally as programming puzzles [Schuster et al., 2021], a code-based problem format where solutions can easily be verified for correctness by execution. In experiments on publicly-available LMs, test accuracy more than doubles. This work demonstrates the potential for code LMs, with an interpreter, to generate instructive problems and improve their own performance.

cs.LG↗

arXiv2020

Representing Partial Programs with Blended Abstract Semantics

Synthesizing programs from examples requires searching over a vast, combinatorial space of possible programs. In this search process, a key challenge is representing the behavior of a partially written program before it can be executed, to judge if it is on the right track and predict where to search next. We introduce a general technique for representing partially written programs in a program synthesis engine. We take inspiration from the technique of abstract interpretation, in which an approximate execution model is used to determine if an unfinished program will eventually satisfy a goal specification. Here we learn an approximate execution model implemented as a modular neural network. By constructing compositional program representations that implicitly encode the interpretation semantics of the underlying programming language, we can represent partial programs using a flexible combination of concrete execution state and learned neural representations, using the learned approximate semantics when concrete semantics are not known (in unfinished parts of the program). We show that these hybrid neuro-symbolic representations enable execution-guided synthesizers to use more powerful language constructs, such as loops and higher-order functions, and can be used to synthesize programs more accurately for a given search budget than pure neural approaches in several domains.

cs.PL↗

arXiv2016

Activity measurement of 60Fe through the decay of 60mCo and confirmation of its half-life

The half-life of the neutron-rich nuclide, {\fesixty} has been in dispute in recent years. A measurement in 2009 published a value of $(2.62 \pm 0.04)\times10^{6}$ years, almost twice that of the previously accepted value from 1984 of $(1.49 \pm 0.27)\times10^{6}$ years. This longer half-life was confirmed in 2015 by a second measurement, resulting in a value of $(2.50 \pm 0.12)\times10^{6}$ years. All three half-life measurements used the grow-in of the $\gamma$-ray lines in {\nisixty} from the decay of the ground state of $^{60}\text{Co}$ (t$_{1/2}$=5.27 years) to determine the activity of a sample with a known number of {\fesixty} atoms. In contrast, the work presented here measured the {\fesixty} activity directly via the 58.6 keV $\gamma$-ray line from the short-lived isomeric state of $^{60}\text{Co}$ (t$_{1/2}$=10.5 minutes), thus being independent of any possible contamination from long-lived $^{60\text{g}}\text{Co}$. A fraction of the material from the 2015 experiment with a known number of {\fesixty} atoms was used for the activity measurement, resulting in a half-life value of $(2.72 \pm 0.16)\times10^{6}$ years, confirming again the longer half-life. In addition, {\fesixty}/{\fe} isotopic ratios of samples with two different dilutions of this material were measured with Accelerator Mass Spectrometry (AMS) to determine the number of {\fesixty} atoms. Combining this with our activity measurement resulted in a half-life value of $(2.69 \pm 0.28)\times 10^{6}$ years, again agreeing with the longer half-life.

nucl-ex↗

arXiv2013

Measurement of the 33S(α,p)36Cl cross section: Implications for production of 36Cl in the early Solar System

Short-lived radionuclides (SLRs) with lifetimes τ< 100 Ma are known to have been extant when the Solar System formed over 4.5 billion years ago. Identifying the sources of SLRs is important for understanding the timescales of Solar System formation and processes that occurred early in its history. Extinct 36Cl (t_1/2 = 0.301 Ma) is thought to have been produced by interaction of solar energetic particles (SEPs), emitted by the young Sun, with gas and dust in the nascent Solar System. However, models that calculate SLR production in the early Solar System (ESS) lack experimental data for the 36Cl production reactions. We present here the first measurement of the cross section of one of the main 36Cl production reactions, 33S(α,p)36Cl, in the energy range 0.70 - 2.42 MeV/A. The cross section measurement was performed by bombarding a target and collecting the recoiled 36Cl atoms produced in the reaction, chemically processing the samples, and measuring the 36Cl/Cl ratio of the activated samples with accelerator mass spectrometry (AMS). The experimental results were found to be systematically higher than the cross sections used in previous local irradiation models and other Hauser-Feshbach calculated predictions. However, the effects of the experimentally measured cross sections on the modeled production of 36Cl in the early Solar System were found to be minimal. Reactions channels involving S targets dominate 36Cl production, but the astrophysical event parameters can dramatically change each reactions' relative contribution.

nucl-ex↗