Self-adjusting Computation Research Articles

Vision Transformer With Enhanced Self-Attention for Few-Shot Ship Target Recognition in Complex Environments

Ship target recognition is essential for maritime transportation, commercial trade, maritime security, and monitoring illegal activity. The majority of previous ship target recognition models are based on fully-supervised learning, which necessitates vast amounts of labeled training data. However, optical lenses capturing ship images in natural scenarios have numerous complex environments, making feature extraction difficult due to hardware, weather, light, waves, and other factors. Moreover, image annotation is also both expensive and time-consuming. How to effectively learn from unbalanced, limited, and sparsely annotated sample data is a significant challenge. In this paper, we propose a new vision transformer with enhanced self-attention(SAVT) that can locate ship targets in complex environments with just a few shots. SAVT consists of a novel enhanced self-attention(ESA) module and a novel measurement self-adjusting computation(MAC) module to differentiate unique ship classes in complex environments with few labeled samples. The ESA module focuses on the features of the target-region at each layer of the backbone network in order to ensure that the contour detailed features are extracted effectively from ship images with complex environments. The model is then enhanced in terms of speed and stability by utilizing the MAC module. CIB-ships and public MAR-ships datasets are used to evaluate the proposed method. In this paper, the results demonstrate that SAVT achieves good performance on few-shot labeled data under both 1-shot and 5-shot settings, while also providing valuable guidance for maritime ship target recognition with limited labeled samples.

Functional programming for dynamic and large data with self-adjusting computation

Combining type theory, language design, and empirical work, we present techniques for computing with large and dynamically changing datasets. Based on lambda calculus, our techniques are suitable for expressing a diverse set of algorithms on large datasets and, via self-adjusting computation, enable computations to respond automatically to changes in their data. To improve the scalability of self-adjusting computation, we present a type system for precise dependency tracking that minimizes the time and space for storing dependency metadata. The type system eliminates an important assumption of prior work that can lead to recording spurious dependencies. We present a type-directed translation algorithm that generates correct self-adjusting programs without relying on this assumption. We then show a probabilistic-chunking technique to further decrease space usage by controlling the fundamental space-time tradeoff in self-adjusting computation. We implement and evaluate these techniques, showing promising results on challenging benchmarks involving large graphs.

Implicit self-adjusting computation for purely functional programs

AbstractComputational problems that involve dynamic data, such as physics simulations and program development environments, have been an important subject of study in programming languages. Building on this work, recent advances in self-adjusting computation have developed techniques that enable programs to respond automatically and efficiently to dynamic changes in their inputs. Self-adjusting programs have been shown to be efficient for a reasonably broad range of problems, but the approach still requires an explicit programming style, where the programmer must use specific monadic types and primitives to identify, create, and operate on data that can change over time. We describe techniques for automatically translating purely functional programs into self-adjusting programs. In this implicit approach, the programmer need only annotate the (top-level) input types of the programs to be translated. Type inference finds all other types, and a type-directed translation rewrites the source program into an explicitly self-adjusting target program. The type system is related to information-flow type systems and enjoys decidable type inference via constraint solving. We prove that the translation outputs well- typed self-adjusting programs and preserves the source program's input–output behavior, guaranteeing that translated programs respond correctly to all changes to their data. Using a cost semantics, we also prove that the translation preserves the asymptotic complexity of the source program.

A consistent semantics of self-adjusting computation

AbstractThis paper presents a semantics of self-adjusting computation and proves that the semantics is correct and consistent. The semantics introduces memoizing change propagation, which enhances change propagation with the classic idea of memoization to enable reuse of computations even when memory is mutated via side effects. During evaluation, computation reuse via memoization triggers a change-propagation algorithm that adapts the reused computation to the memory mutations (side effects) that took place since the creation of the computation. Since the semantics includes both memoization and change propagation, it involves both non-determinism (due to memoization) and mutation (due to change propagation). Our consistency theorem states that the non-determinism is not harmful: any two evaluations of the same program starting at the same state yield the same result. Our correctness theorem states that mutation is not harmful: Self-adjusting programs are compatible with purely functional programming. We formalize the semantics and its meta-theory in the LF logical framework and machine check our proofs using Twelf.

Type-directed automatic incrementalization

Application data often changes slowly or incrementally over time. Since incremental changes to input often result in only small changes in output, it is often feasible to respond to such changes asymptotically more efficiently than by re-running the whole computation. Traditionally, realizing such asymptotic efficiency improvements requires designing problem-specific algorithms known as dynamic or incremental algorithms, which are often significantly more complicated than conventional algorithms to design, analyze, implement, and use. A long-standing open problem is to develop techniques that automatically transform conventional programs so that they correctly and efficiently respond to incremental changes. In this paper, we describe a significant step towards solving the problem of automatic incrementalization: a programming language and a compiler that can, given a few type annotations describing what can change over time, compile a conventional program that assumes its data to be static (unchanging over time) to an incremental program. Based on recent advances in self-adjusting computation, including a theoretical proposal for translating purely functional programs to self-adjusting programs, we develop techniques for translating conventional Standard ML programs to self-adjusting programs. By extending the Standard ML language, we design a fully featured programming language with higher-order features, a module system, and a powerful type system, and implement a compiler for this language. The resulting programming language, LML, enables translating conventional programs decorated with simple type annotations into incremental programs that can respond to changes in their data correctly and efficiently. We evaluate the effectiveness of our approach by considering a range of benchmarks involving lists, vectors, and matrices, as well as a ray tracer. For these benchmarks, our compiler incrementalizes existing code with only trivial amounts of annotation. The resulting programs are often asymptotically more efficient, leading to orders of magnitude speedups in practice.

Self-adjusting stack machines

Self-adjusting computation offers a language-based approach to writing programs that automatically respond to dynamically changing data. Recent work made significant progress in developing sound semantics and associated implementations of self-adjusting computation for high-level, functional languages. These techniques, however, do not address issues that arise for low-level languages, i.e., stack-based imperative languages that lack strong type systems and automatic memory management. In this paper, we describe techniques for self-adjusting computation which are suitable for low-level languages. Necessarily, we take a different approach than previous work: instead of starting with a high-level language with additional primitives to support self-adjusting computation, we start with a low-level intermediate language, whose semantics is given by a stack-based abstract machine. We prove that this semantics is sound: it always updates computations in a way that is consistent with full reevaluation. We give a compiler and runtime system for the intermediate language used by our abstract machine. We present an empirical evaluation that shows that our approach is efficient in practice, and performs favorably compared to prior proposals.

Two for the price of one

Parallel or incremental versions of an algorithm can significantly outperform their counterparts, but are often difficult to develop. Programming models that provide appropriate abstractions to decompose data and tasks can simplify parallelization. We show in this work that the same abstractions can enable both parallel and incremental execution. We present a novel algorithm for parallel self-adjusting computation. This algorithm extends a deterministic parallel programming model (concurrent revisions) with support for recording and repeating computations. On record, we construct a dynamic dependence graph of the parallel computation. On repeat, we reexecute only parts whose dependencies have changed. We implement and evaluate our idea by studying five example programs, including a realistic multi-pass CSS layout algorithm. We describe programming techniques that proved particularly useful to improve the performance of self-adjustment in practice. Our final results show significant speedups on all examples (up to 37x on an 8-core machine). These speedups are well beyond what can be achieved by parallelization alone, while requiring a comparable effort by the programmer.

Implicit self-adjusting computation for purely functional programs

Computational problems that involve dynamic data, such as physics simulations and program development environments, have been an important subject of study in programming languages. Building on this work, recent advances in self-adjusting computation have developed techniques that enable programs to respond automatically and efficiently to dynamic changes in their inputs. Self-adjusting programs have been shown to be efficient for a reasonably broad range of problems but the approach still requires an explicit programming style, where the programmer must use specific monadic types and primitives to identify, create and operate on data that can change over time. We describe techniques for automatically translating purely functional programs into self-adjusting programs. In this implicit approach, the programmer need only annotate the (top-level) input types of the programs to be translated. Type inference finds all other types, and a type-directed translation rewrites the source program into an explicitly self-adjusting target program. The type system is related to information-flow type systems and enjoys decidable type inference via constraint solving. We prove that the translation outputs well-typed self-adjusting programs and preserves the source program's input-output behavior, guaranteeing that translated programs respond correctly to all changes to their data. Using a cost semantics, we also prove that the translation preserves the asymptotic complexity of the source program.

Traceable data types for self-adjusting computation

Self-adjusting computation provides an evaluation model where computations can respond automatically to modifications to their data by using a mechanism for propagating modifications through the computation. Current approaches to self-adjusting computation guarantee correctness by recording dependencies in a trace at the granularity of individual memory operations. Tracing at the granularity of memory operations, however, has some limitations: it can be asymptotically inefficient (\eg, compared to optimal solutions) because it cannot take advantage of problem-specific structure, it requires keeping a large computation trace (often proportional to the runtime of the program on the current input), and it introduces moderately large constant factors in practice. In this paper, we extend dependence-tracing to work at the granularity of the query and update operations of arbitrary (abstract) data types, instead of just reads and writes on memory cells. This can significantly reduce the number of dependencies that need to be kept in the trace and followed during an update. We define an interface for supporting a traceable version of a data type, which reports the earliest query that depends on (is changed by) revising operations back in time, and implement several such structures, including priority queues, queues, dictionaries, and counters. We develop a semantics for tracing, extend an existing self-adjusting language, ΔML, and its implementation to support traceable data types, and present an experimental evaluation by considering a number of benchmarks. Our experiments show dramatic improvements on space and time, sometimes by as much as two orders of magnitude.

An experimental analysis of self-adjusting computation

Recent work on adaptive functional programming (AFP) developed techniques for writing programs that can respond to modifications to their data by performing change propagation . To achieve this, executions of programs are represented with dynamic dependence graphs (DDGs) that record data dependences and control dependences in a way that a change-propagation algorithm can update the computation as if the program were from scratch, by re-executing only the parts of the computation affected by the changes. Since change-propagation only re-executes parts of the computation, it can respond to certain incremental modifications asymptotically faster than recomputing from scratch, potentially offering significant speedups. Such asymptotic speedups, however, are rare: for many computations and modifications, change propagation is no faster than recomputing from scratch. In this article, we realize a duality between dynamic dependence graphs and memoization, and combine them to give a change-propagation algorithm that can dramatically increase computation reuse. The key idea is to use DDGs to identify and re-execute the parts of the computation that are affected by modifications, while using memoization to identify the parts of the computation that remain unaffected by the changes. We refer to this approach as self-adjusting computation. Since DDGs are imperative, but (traditional) memoization requires purely functional computation, reusing computation correctly via memoization becomes a challenge. We overcome this challenge with a technique for remembering and reusing not just the results of function calls (as in conventional memoization), but their executions represented with DDGs. We show that the proposed approach is realistic by describing a library for self-adjusting computation, presenting efficient algorithms for realizing the library, and describing and evaluating an implementation. Our experimental evaluation with a variety of applications, ranging from simple list primitives to more sophisticated computational geometry algorithms, shows that the approach is effective in practice: compared to recomputing from-scratch; self-adjusting programs respond to small modifications to their data orders of magnitude faster.

CEAL

Self-adjusting computation offers a language-centric approach to writing programs that can automatically respond to modifications to their data (e.g., inputs). Except for several domain-specific implementations, however, all previous implementations of self-adjusting computation assume mostly functional, higher-order languages such as Standard ML. Prior to this work, it was not known if self-adjusting computation can be made to work with low-level, imperative languages such as C without placing undue burden on the programmer. We describe the design and implementation of CEAL: a C-based language for self-adjusting computation. The language is fully general and extends C with a small number of primitives to enable writing self-adjusting programs in a style similar to conventional C programs. We present efficient compilation techniques for translating CEAL programs into C that can be compiled with existing C compilers using primitives supplied by a run-time library for self-adjusting computation. We implement the proposed compiler and evaluate its effectiveness. Our experiments show that CEAL is effective in practice: compiled self-adjusting programs respond to small modifications to their data by orders of magnitude faster than recomputing from scratch while slowing down a from-scratch run by a moderate constant factor. Compared to previous work, we measure significant space and time improvements.

A cost semantics for self-adjusting computation

Self-adjusting computation is an evaluation model in which programs can respond efficiently to small changes to their input data by using a change-propagation mechanism that updates computation by re-building only the parts affected by changes. Previous work has proposed language techniques for self-adjusting computation and showed the approach to be effective in a number of application areas. However, due to the complex semantics of change propagation and the indirect nature of previously proposed language techniques, it remains difficult to reason about the efficiency of self-adjusting programs and change propagation. In this paper, we propose a cost semantics for self-adjusting computation that enables reasoning about its effectiveness. As our source language, we consider a direct-style λ-calculus with first-class mutable references and develop a notion of trace distance for source programs. To facilitate asymptotic analysis, we propose techniques for composing and generalizing concrete distances via trace contexts (traces with holes). We then show how to translate the source language into a self-adjusting target language such that the translation (1) preserves the extensional semantics of the source programs and the cost of from-scratch runs, and (2) ensures that change propagation between two evaluations takes time bounded by their relative distance. We consider several examples and analyze their effectiveness by considering upper and lower bounds.

Compiling self-adjusting programs with continuations

Self-adjusting programs respond automatically and efficiently to input changes by tracking the dynamic data dependences of the computation and incrementally updating the output as needed. In order to identify data dependences, previously proposed approaches require the user to make use of a set of monadic primitives. Rewriting an ordinary program into a self-adjusting program with these primitives, however, can be difficult and error-prone due to various monadic and proper-usage restrictions, some of which cannot be enforced statically. Previous work therefore suggests that self-adjusting computation would benefit from direct language and compiler support. In this paper, we propose a language-based technique for writing and compiling self-adjusting programs from ordinary programs. To compile self-adjusting programs, we use a continuation-passing style (cps) transformation to automatically infer a conservative approximation of the dynamic data dependences. To prevent the inferred, approximate dependences from degrading the performance of change propagation, we generate memoized versions of cps functions that can reuse previous work even when they are invoked with different continuations. The approach offers a natural programming style that requires minimal changes to existing code, while statically enforcing the invariants required by self-adjusting computation. We validate the feasibility of our proposal by extending Standard ML and by integrating the transformation into MLton, a whole-program optimizing compiler for Standard ML. Our experiments indicate that the proposed compilation technique can produce self-adjusting programs whose performance is consistent with the asymptotic bounds and experimental results obtained via manual rewriting (up to a constant factor).

Imperative self-adjusting computation

Self-adjusting computation enables writing programs that can automatically and efficiently respond to changes to their data (e.g., inputs). The idea behind the approach is to store all data that can change over time in modifiable references and to let computations construct traces that can drive change propagation. After changes have occurred, change propagation updates the result of the computation by re-evaluating only those expressions that depend on the changed data. Previous approaches to self-adjusting computation require that modifiable references be written at most once during execution---this makes the model applicable only in a purely functional setting. In this paper, we present techniques for imperative self-adjusting computation where modifiable references can be written multiple times. We define a language SAIL (Self-Adjusting Imperative Language) and prove consistency, i.e., that change propagation and from-scratch execution are observationally equivalent. Since SAIL programs are imperative, they can create cyclic data structures. To prove equivalence in the presence of cycles in the store, we formulate and use an untyped, step-indexed logical relation, where step indices are used to ensure well-foundedness. We show that SAIL accepts an asymptotically efficient implementation by presenting algorithms and data structures for its implementation. When the number of operations (reads and writes) per modifiable is bounded by a constant, we show that change propagation becomes as efficient as in the non-imperative case. The general case incurs a slowdown that is logarithmic in the maximum number of such operations. We describe a prototype implementation of SAIL as a Standard ML library.

An experimental analysis of self-adjusting computation

Dependence graphs and memoization can be used to efficiently update the output of a program as the input changes dynamically. Recent work has studied techniques for combining these approaches to effectively dynamize a wide range of applications. Toward this end various theoretical results were given. In this paper we describe the implementation of a library based on these ideas, and present experimental results on the efficiency of this library on a variety of applications. The results of the experiments indicate that the approach is effective in practice, often requiring orders of magnitude less time than recomputing the output from scratch. We believe this is the first experimental evidence that incremental computation of any type is effective in practice for a reasonably broad set of applications.

A Library for Self-Adjusting Computation

We present a Standard ML library for writing programs that automatically adjust to changes to their data. The library combines modifiable references and memoization to achieve efficient updates. We describe an implementation of the library and apply it to the problem of maintaining the convex hull of a dynamically changing set of points. Our experiments show that the overhead of the library is small, and that self-adjusting programs can adjust to small changes three-orders of magnitude faster than recomputing from scratch. The implementation relies on invariants that could be enforced by a modal type system. We show, using an existing language, abstract interfaces for modifiable references and for memoization that ensure the same safety properties without the use of modal types. The interface for memoization, however, does not scale well, suggesting a language-based approach to be preferable after all.