摘要 :
Automatically transforming developers' natural language descriptions into source code has been a longstanding goal in software engineering research. Two types of approaches have been proposed in the literature to achieve this: cod...
展开
Automatically transforming developers' natural language descriptions into source code has been a longstanding goal in software engineering research. Two types of approaches have been proposed in the literature to achieve this: code generation, which involves generating a new code snippet, and code search, which involves reusing existing code. However, despite existing efforts, the effectiveness of the state-of-the-art techniques remains limited. To seek for further advancement, our insight is that code generation and code search can help overcome the limitation of each other: the code generator can benefit from feedback on the quality of its generated code, which can be provided by the code searcher, while the code searcher can benefit from the additional training data augmented by the code generator to better understand code semantics. Drawing on this insight, we propose a novel approach that combines code generation and code search techniques using a generative adversarial network (GAN), enabling mutual improvement through the adversarial training. Specifically, we treat code generation and code search as the generator and discriminator in the GAN framework, respectively, and incorporate several customized designs for our tasks. We evaluate our approach in eight different settings, and consistently observe significant performance improvements for both code generation and code search. For instance, when using NatGen, a state-of-the-art code generator, as the generator and GraphCodeBERT, a state-of-the-art code searcher, as the discriminator, we achieve a 32% increase in CodeBLEU score for code generation, and a 12% increase in mean reciprocal rank for code search on a large-scale Python dataset, compared to their original performances.
收起
摘要 :
This article presents a design for the Denali-2 superoptimizer, which will generate minimum-instruction-length machine code for realistic machine architectures using automatic theorem-proving technology: specifically, using E-grap...
展开
This article presents a design for the Denali-2 superoptimizer, which will generate minimum-instruction-length machine code for realistic machine architectures using automatic theorem-proving technology: specifically, using E-graph matching (a technique for pattern matching in the presence of equality information) and Boolean satisfiability solving. This article presents a precise definition of the underlying automatic programming problem solved by the Denali-2 superoptimizer. It sketches the E-graph matching phase and presents a detailed exposition and proof of soundness of the reduction of the automatic programming problem to the Boolean satisfiability problem.
收起
摘要 :
We propose a mew technique for constructing code-generator generators, which combines the advantages of the Graham-Glanville parsing technique and the bottom-up tree parsing approach. Machine descriptions are similar to Yacc speci...
展开
We propose a mew technique for constructing code-generator generators, which combines the advantages of the Graham-Glanville parsing technique and the bottom-up tree parsing approach. Machine descriptions are similar to Yacc specifications. The construction effectively generates a Pushdown automaton as the matching device. This device is able to handle ambiguous grammars, And can be used to generate locally optimal code without the use of heuristics.
收起
摘要 :
NGUYEN et al. give a method for constructing binary CP constant-weight (CPC) codes, then they construct protocol sequences with good performance by using the codes. To make the minimum distance of the CP codes large enough, Nguyen...
展开
NGUYEN et al. give a method for constructing binary CP constant-weight (CPC) codes, then they construct protocol sequences with good performance by using the codes. To make the minimum distance of the CP codes large enough, Nguyen et al. select two classes of maximum distance separable (MDS) codes which are q-ary Reed-Solomon (R-S) codes and q-ary generalized Berlekamp-Justesen (B-J) codes. The generalized B-J codes are more interesting because of their larger code length and minimum distance. It canbe seen from ref. that the construction from R-S codes is asymptotically optimum, but the construction from generalized B-J codes is asymptotically bad since its efficiency is only about 1/g which turns out to be 0 when q turns out to be infinity. Hence, how to improve the efficiency of the construction for generalized B-J codes is a key problem, In this note the period distribution of generalized B-J codes is proposed. Then, we give a method of constructing asymptotically optimum CP codes and generalized CP codes. Thus, a class of more efficient binary CPC codes and protocol sequences for collision channel without feedback can be obtained.
收起
摘要 :
Let R = F-q + uF(q), where q is a power of a prime number p and u(2) = 0. A triple cyclic code of length (r; s; t) over R is a set that can be partitioned into three parts that any cyclic shift of the coordinates of the three part...
展开
Let R = F-q + uF(q), where q is a power of a prime number p and u(2) = 0. A triple cyclic code of length (r; s; t) over R is a set that can be partitioned into three parts that any cyclic shift of the coordinates of the three parts leaves the code invariant. These codes can be viewed as R[x]-submodules of R[x]/< x(r) - 1 > R[x]=< x(s) - 1 > x R[x]/ < x(t) - 1 >. In this paper, we study the generator polynomials and the minimum generating sets of this kind of codes. Some optimal or almost optimal linear codes are obtained from this family of codes. We present the relationship between the generators of triple cyclic codes and their duals. As a special class of triple cyclic codes, separable codes over R are discussed briefly in the end.
收起
摘要 :
In this paper a general class of linear cyclic codes , is defined of length and over a field with . This class of codes includes as special cases quadratic residue codes, generalized quadratic residue codes, -residue codes and -co...
展开
In this paper a general class of linear cyclic codes , is defined of length and over a field with . This class of codes includes as special cases quadratic residue codes, generalized quadratic residue codes, -residue codes and -codes. Furthermore, they partially overlap with the families of duadic, triadic and polyadic codes. Expressions for idempotent generators are derived in terms of the size of cyclotomic cosets mod and coefficients of the irreducible polynomials over dividing . As an auxiliary tool an orthonormal matrix is introduced whose columns correspond to these idempotents. Concrete examples are presented for and , where is an arbitrary odd prime. When or the codes all belong to the subclass of 2-residue codes. Using this technique, we determine the idempotents of the codes , and recover those of the generalized quadratic codes and of the codes . In the final section the idempotents of the cubic residue codes are constructed.
收起
摘要 :
In this paper, we propose a mechanism on how to construct
long MDS self-dual codes from short ones. These codes are
special types of generalized Reed-Solomon (GRS) codes or
extended generalized Reed-Solomon codes. The main tool...
展开
In this paper, we propose a mechanism on how to construct
long MDS self-dual codes from short ones. These codes are
special types of generalized Reed-Solomon (GRS) codes or
extended generalized Reed-Solomon codes. The main tool is
utilizing additive structure or multiplicative structure on finite
fields. By applying this method, more MDS self-dual codes can
be constructed.
收起
摘要 :
Whereas Huffman coding finds a prefix code minimizing mean codeword length for a given finite-item probability distribution, quasiarithmetic or quasilinear coding problems have the goal of minimizing a generalized mean of the form...
展开
Whereas Huffman coding finds a prefix code minimizing mean codeword length for a given finite-item probability distribution, quasiarithmetic or quasilinear coding problems have the goal of minimizing a generalized mean of the form$varphi ^-1(sum _i p_ivarphi (l_i))$, where$l_i$denotes the length of the$i$th codeword,$p_i$denotes the corresponding probability, and$varphi$is a monotonically increasing cost function. Such problems, proposed by Campbell, have a number of diverse applications. Several cost functions are shown here to yield quasiarithmetic problems with simple redundancy bounds in terms of a generalized entropy. A related property, also shown here, involves the existence of optimal codes: For “well-behaved” cost functions, optimal codes always exist for (possibly infinite-alphabet) sources having finite generalized entropy. An algorithm is introduced for finding binary codes optimal for convex cost functions. This algorithm, which can be extended to other minimization utilities, can be performed using quadratic time and linear space. This reduces the computational complexity of a problem involving minimum delay in a queue, allows combinations of previously considered problems to be optimized, and greatly expands the set of problems solvable in quadratic time and linear space.
收起
摘要 :
A double cyclic code of length (r, s) over a chain ring R is a set that can be partitioned into two parts that any cyclic shift of the coordinates of both parts leaves invariant the code. These codes can be viewed as R[x]-submodul...
展开
A double cyclic code of length (r, s) over a chain ring R is a set that can be partitioned into two parts that any cyclic shift of the coordinates of both parts leaves invariant the code. These codes can be viewed as R[x]-submodules of R~r × R~s. In this paper, the generator polynomials of this family of codes as R[x]-submodules of R~r × Rs are determined. Further, the minimal generating sets of this family of codes as R-submodules of R~r × R~s are obtained. Finally, we show the relationship of generators between the double cyclic code and its dual.
收起
摘要 :
In this paper, we study quadratic residue (QR) codes of prime length q over the ring R = F-p + uF(p) + vF(p) with u(2) = u, v(2) = v and uv = vu = 0, where p and q are distinct odd prime numbers. We analyze some basic properties o...
展开
In this paper, we study quadratic residue (QR) codes of prime length q over the ring R = F-p + uF(p) + vF(p) with u(2) = u, v(2) = v and uv = vu = 0, where p and q are distinct odd prime numbers. We analyze some basic properties of cyclic codes of length n over R, we define QR codes by their generating idempotents. Further, we discuss the extended QR codes. We present a considerable number of good p-ary codes as Gray images of QR codes over F-p + uF(p) + vF(p) by considering the case when p is an odd prime.
收起