AIPLANS (Advances in Programming Languages and Neurosymbolic Systems): a new workshop at NeurIPS 2021 fusing ML with programming theory to create neurosymbolic program-writing machines!
- Paper submission deadline: Oct. 4th, 2021 AoE
- Reviews released: Oct. 23rd, 2021
- Camera-ready deadline: Dec. 10, 2021 AoE
- Tentative conference dates: Dec. 14th, 2021
Our workshop brings together researchers from various backgrounds. We believe developing neurosymbolic systems will require engineers, designers and theorists from statistical learning and programming language research.
- Machine learning researchers can present advances in meta-learning, reinforcement learning and program synthesis. AIPLANS offers these participants an opportunity to share their research and learn about new automatic programming languages and techniques for inference.
- Programming language designers can give insight into the design and implementation of automatic programming languages and DSLs. AIPLANS offers them the opportunity to gather feedback about design choices, promote the language and engage with their users.
- Programming language theorists can present fundamental theory of mechanical reasoning and automatic programming languages, such as functional, semiring or array programming. AIPLANS will help them bridge the gap between theory and practice, and gain insight into the capabilities and limitations of machine learning technology.
We would be excited to see submissions similar to or building on the following non-exhaustive list of topics:
- Algorithms for automatic differentiation and inference in ML systems:
- Design of tools for automatic differentiation:
- Development of neurosymbolic based reasoning systems, e.g.:
- Program synthesis:
- Improvements to RobustFill, DeepCoder or applications such as Naturalizing A Programming Language Via Interactive Learning
- Probabilistic Programming:
- Theory of mechanical reasoning and automatic programming languages:
- Work which explains the interplay between neural networks and programming methods, e.g., Dynamical systems that sort lists, diagonalize matrices and solve linear programming problems, Thinking Like Transformers
- Probabilistic programming frameworks, e.g., Stan, Edward, PyMC3, Pyro, torch-struct
- Functional or array programming, e.g., Hasktorch, Strongly-Typed RNNs, Tangent
- Declarative programming / constraint programming
- e.g. Neural program synthesis, neural guided program search
- Dynamic programming / reinforcement learning
- Functional programming / λ-calculus
- Array programming / linear algebra
- Logic programming / Relational programming
- Prolog, Datalog, miniKanren, HOL, LF/Twelf
- Computer aided reasoning / automatic theorem proving
- Domain-specific languages and compilers
- Inductive programming / programming by example
Developers of languages, frameworks and libraries, including those who traditionally publish in venues such as SIGPLAN and SIGSOFT are encouraged to consider submitting ongoing work that may be relevant to machine learning community. Details regarding evaluation criteria, deadlines and workshop logistics can be found under the AIPLANS CFP.
AIPLANS is brought to you in collaboration with the organizers of the Differentiable Programming Workshop at NeurIPS 2021. We share their enthusiasm for differentiable and probabilistic programming and see many applications towards program synthesis and symbolic reasoning. Those with similar interests are highly encouraged to participate in both workshops. Prospective authors should submit their work to the most relevant venue, as dual submissions will be evaluated once.