Anthony Di Franco
2015-10-25 01:25:05 UTC
Hi all,
Sorry about the late notice, but it very recently occurred to me that this
would be a good group to invite feedback from on a talk I'm giving this
coming Monday on approximate search techniques for problems in programming
languages. It's my attempt to describe for the programming language
community the broader agenda from this
<http://www.meetup.com/Bay-Area-entrepreneur-in-statistics/events/224830306/>
talk I gave last month, which
covered, I suppose I could say, the intersection of nondeterminism and
computation in a very broad interdisciplinary way. To help me with this
shift in perspective, I could use some feedback on whether this makes sense
and how I might improve the focus, either before the talk or after. What
I'm trying to move towards is, in one sense, a generalization of constraint
solving to arbitrary recursive relationships potentially involving
uncertainty, so there's a lot of overlap in the applications with VPRI-like
work and goals, such as declarative UI, better, easier parsing, and Hesam
Samimi's PlanB, for example.
Hope to see some of you there or otherwise hear your thoughts. Here's the
info:
http://www.meetup.com/SF-Types-Theorems-and-Programming-Languages/events/226067416/
*Finally Fifth? Searching for answers in an uncertain world.*
Monday, October 26, 2015, 7:00 PM
Mixrank
<http://maps.google.com/maps?f=q&hl=en&q=164+Townsend+St.+%234%2C+San+Francisco%2C+CA%2C+94107%2C+us>,
164
Townsend St. #4, San Francisco, CA
*Most of us are probably familiar with the trajectory software projects
take: quick early progress with few people working on them, which
transitions, as the scope grows, continuously but sharply to a regime where
large numbers of people and large amounts of effort, at the scale of some
of the largest corporations in history, are insufficient even to keep up
with already known problems. Typically, motivated by the prevailing formal
logic background in programming language theory, people turn to software
methodologies with stronger a priori guarantees to mitigate this problem,
such as functional programming with types, but I will propose a different,
though not mutually exclusive, approach, drawing on a control theory and
systems theory background. Motivated by Robert Kowalski's perspective
developed in his "Algorithm = Logic + Control" (1979) I claim that the real
problem is the combinatorial explosion in the number of algorithms required
to enforce a desired set of relationships as that set of relationships
grows in size. The solution is to finally come to grips with
nondeterminism, and the solution to that, in turn, is to use approximate
search techniques that can take advantage of uncertain information,
information feedback, and compression of the search space. This motivates
the design of the "Fifth" software system I'm currently working on. We'll
conclude with a description of work in progress on the Fifth system.*
Sorry about the late notice, but it very recently occurred to me that this
would be a good group to invite feedback from on a talk I'm giving this
coming Monday on approximate search techniques for problems in programming
languages. It's my attempt to describe for the programming language
community the broader agenda from this
<http://www.meetup.com/Bay-Area-entrepreneur-in-statistics/events/224830306/>
talk I gave last month, which
covered, I suppose I could say, the intersection of nondeterminism and
computation in a very broad interdisciplinary way. To help me with this
shift in perspective, I could use some feedback on whether this makes sense
and how I might improve the focus, either before the talk or after. What
I'm trying to move towards is, in one sense, a generalization of constraint
solving to arbitrary recursive relationships potentially involving
uncertainty, so there's a lot of overlap in the applications with VPRI-like
work and goals, such as declarative UI, better, easier parsing, and Hesam
Samimi's PlanB, for example.
Hope to see some of you there or otherwise hear your thoughts. Here's the
info:
http://www.meetup.com/SF-Types-Theorems-and-Programming-Languages/events/226067416/
*Finally Fifth? Searching for answers in an uncertain world.*
Monday, October 26, 2015, 7:00 PM
Mixrank
<http://maps.google.com/maps?f=q&hl=en&q=164+Townsend+St.+%234%2C+San+Francisco%2C+CA%2C+94107%2C+us>,
164
Townsend St. #4, San Francisco, CA
*Most of us are probably familiar with the trajectory software projects
take: quick early progress with few people working on them, which
transitions, as the scope grows, continuously but sharply to a regime where
large numbers of people and large amounts of effort, at the scale of some
of the largest corporations in history, are insufficient even to keep up
with already known problems. Typically, motivated by the prevailing formal
logic background in programming language theory, people turn to software
methodologies with stronger a priori guarantees to mitigate this problem,
such as functional programming with types, but I will propose a different,
though not mutually exclusive, approach, drawing on a control theory and
systems theory background. Motivated by Robert Kowalski's perspective
developed in his "Algorithm = Logic + Control" (1979) I claim that the real
problem is the combinatorial explosion in the number of algorithms required
to enforce a desired set of relationships as that set of relationships
grows in size. The solution is to finally come to grips with
nondeterminism, and the solution to that, in turn, is to use approximate
search techniques that can take advantage of uncertain information,
information feedback, and compression of the search space. This motivates
the design of the "Fifth" software system I'm currently working on. We'll
conclude with a description of work in progress on the Fifth system.*