Josh Fass

`periodic_torsion` becomes undefined for collinear inputs, and the associated force magnitude becomes large for nearly collinear inputs. Such configurations are extremely rare in typical simulations, due to large force constants...

After upgrading, can revert https://github.com/proteneer/timemachine/pull/829

Migrated from https://github.com/proteneer/timemachine/pull/825#discussion_r970037993 Ideally thermodynamic parameters should be as consistent as possible between simulation conditions and experimental conditions, but (1) small differences may be inconsequential, (2) there may be overriding...

Migrated from https://github.com/proteneer/timemachine/pull/738 At least four approaches possible: * Current approach -- stereo unaware, but occasionally introducing very small asymmetries in cases like benzene * Initial approach proposed in #738...

MBAR uncertainty estimates in a low-overlap case

2

Here is an example that came up in a discussion with @mrshirts , where MBAR's uncertainty estimate can revert to a small number when the overlap is brought close to...

Move switch_fn from nonbonded.py to attic. Doesn't have expected boundary conditions when cutoff != 1.0, and is unused (only usages were in commented-out code when introduced in https://github.com/proteneer/timemachine/pull/205 ). Possible...

Consider unifying approach to PRNG state

5

Thanks to @mcwitt for thoughtful comments: migrating from https://github.com/proteneer/timemachine/pull/978#discussion_r1140414877 . Would be good to discuss and adopt project-wide conventions, if possible. Some approaches currently used: * some objects store and...

Migrated from https://github.com/proteneer/timemachine/pull/1119#discussion_r1293482382 * possibly expand test coverage from HarmonicBond to the other Potential classes * possibly test on more random instances

Migrating from PR thread: https://github.com/proteneer/timemachine/pull/949#discussion_r1098786956 -- some changes could potentially be cleaner if Timemachine's Forcefield serialization format more closely matched the structure of Timemachine's Forcefield object. Current: A `.py` file...

Following up on https://github.com/proteneer/timemachine/pull/922#pullrequestreview-1205256185 In principle, per-component decomposition can be misleading. Here is a contrived example: ```python import numpy as np np.random.seed(0) print( """ Problem: Interpreting per-component delta_us can be...