Datalog is a deductive query language for relational databases. We introduce LogiQL, a language based on Datalog and show how it can be used to specify mixedinteger linear optimization models and solve them. Unlike pure algebraic modeling languages, LogiQL allows the user to both specify models, and manipulate and transform the inputs and outputs of the models. This is an advantage over conventional optimization modeling languages that rely on reading data via plug-in tools or importing data from external sources via files. In this chapter, we give a brief overview of LogiQL and describe two mixed integer programming case studies: a production-transportation model and a formulation of the traveling salesman problem.

The standard process of data science tasks is to prepare features inside a database, export them as a denormalized data frame and then apply machine learning algorithms. This process is not optimal for two reasons. First, it requires denormalization of the database that can convert a small data problem into a big data problem. The second shortcoming is that it assumes that the machine learning algorithm is disentangled from the relational model of the problem. That seems to be a serious limitation since the relational model contains very valuable domain expertise. In this paper we explore the use of convex optimization and specifically linear programming, for modelling machine learning algorithms on relational data in an integrated way with data processing operators. We are using SolverBlox, a framework that accepts as an input Datalog code and feeds it into a linear programming solver. We demonstrate the expression of common machine learning algorithms and present use case scenarios where combining data processing with modelling of optimization problems inside a database offers significant advantages.

Abstract : In the probabilistic-programming paradigm, the application logic is specified by means of a description of a probabilistic model (by stating how a sample is being produced) using a Probabilistic Programming Language (PPL). The principal value one obtains from a probabilistic program lies in the inference thereof, that is, reasoning about the entire probability distribution that the program defines (e.g., finding a likely event or estimating its marginal probability). The PPAML kickoff meeting highlighted several research challenges regarding the development of inference infrastructure for PPL, for both increasing software efficiency and reducing software complexity, towards the goal of broadening the PPL applications and the community of implementers and programmers. These challenges include the design of an Application Program Interface (API), or alternatively an Intermediate Representation Language (IRL), that would allow new solvers to be plugged into existing PPLs, and for PPL engines to be able to pick from and combine solvers for a given problem.

The standard process of data science tasks is to prepare features inside a database, export them as a denormalized data frame and then apply machine learning algorithms. This process is not optimal for two reasons. First, it requires denormalization of the database that can convert a small data problem into a big data problem. The second problem is that it assumes that the machine learning algorithm is disentangled from the relational model of the problem. That seems to be a serious limitation since the relational model contains very valuable domain expertise. In this paper we explore the use of convex optimization and specifically linear programming as a data science tool that can express most of the common machine learning algorithms and at the same time it can be natively integrated inside a declarative database. We are using SolverBlox, a framework that accepts as an input Datalog code and feeds it into a linear programming solver. We demonstrate the expression of three common machine learning algorithms, Linear Regression, Factorization Machines and Spectral Clustering, and present use case scenarios where data processing and modelling of optimization problems can be done step by step inside the database.

The LogicBlox system aims to reduce the complexity of software development for modern applications which enhance and automate decision-making and enable their users to evolve their capabilities via a ``self-service'' model. Our perspective in this area is informed by over twenty years of experience building dozens of mission-critical enterprise applications that are in use by hundreds of large enterprises across industries such as retail, telecommunications, banking, and government. We designed and built LogicBlox to be the system we wished we had when developing those applications. In this paper, we discuss the design considerations behind the LogicBlox system and give an overview of its implementation, highlighting innovative aspects. These include: LogiQL, a unified and declarative language based on Datalog; the use of purely functional data structures; novel join processing strategies; advanced incremental maintenance and live programming facilities; a novel concurrency control scheme; and built-in support for prescriptive and predictive analytics.

LogicBlox is a database product designed for enterprise software development, combining transactions and analytics. The underying data model is a relational database, and the query language, LogiQL, is an extension of Datalog [13]. As such, LogiQL features a simple and unified syntax for traditional relational manipulation as well as deeper analytics. Moreover, its declarative nature allows for substantial static analysis for optimizing evaluation schemes, parallelization, and incremental maintenance, and it allows for sophisticated transactional management [11]. In this paper, we describe various extensions of Datalog for supporting prescriptive and predictive analytics. These extensions come in the form of mathematical optimization (mixed integer programming), machine-learning capabilities, statistical relational models, and probabilistic programming. Some of these extensions are currently implemented in LogicQL, while others are in either development or planning phases.

The main aim of the present study was to investigate the effects of soccer-specific training on physical fitness components in adolescent elite soccer players and make comparisons with older counterparts. Twenty two male soccer players from the Serbian First Division team were allocated to two assigned trials according to age – young group (YG) and mature group (MG). Players in their teenage years (19 years and younger) were assigned to YG (10 subjects) and others to MG (12 subjects). Between the first and second test session, all subjects followed six weeks of soccer-specific periodized training programme. There were no differences between groups at preand post-training trial for body mass, vertical jump height, average anaerobic power and VO2max (P>0.05). Body fat was significantly lower in YG before and after training program as compared to MG (P<0.05). Body mass and fat dropped significantly in both groups after training program (P<0.05). Furthermore, average anaerobic power and VO2max along with vertical jump height, were significantly improved in both groups (P<0.05) at posttraining performance. Finally, the magnitude of change in VO2max was significanty superior in MG as compared to YG after training program (18.3 vs. 7.8%; P<0.05). The findings of the present study indicate that the trainability indices are not highly influenced by age in top-level soccer players. (Biol.Sport 26:379-387, 2009)

Bradic, A, Bradic, J, Pasalic, E, and Markovic, G. Isokinetic leg strength profile of elite male basketball players. J Strength Cond Res 23(4): 1332-1337, 2009-In this study, we described the leg strength profile of elite male basketball players and evaluated the positional differences in absolute and relative leg strength. Forty-three elite male basketball players (15 guards, 14 forwards, and 14 centers) performed maximal isokinetic concentric knee extension and flexion efforts (60·s−1 and 180°·s−1) and ankle plantar flexion and dorsiflexion efforts (30°·s−1 and 60°·s−1). Significant (p < 0.01) positional differences in absolute leg muscle strength were observed, with centers having significantly (p < 0.05) greater peak torques compared with guards in all of the tested muscle groups and at all angular velocities. Moreover, centers also possessed significantly (p < 0.05) stronger plantar flexors at 30°·s−1 and dorsiflexors at 60°·s−1 compared with forwards, and forwards possessed significantly (p < 0.05) stronger knee extensors, plantar flexors, and dorsiflexors compared with guards. Normalization of strength values with body mass (N·m·(kg−1)) canceled out the positional differences in knee extensor and flexors strength, but the groups still differed significantly (p < 0.01) in relative plantar flexion and dorsiflexion strength. Specifically, centers had significantly (p < 0.05) stronger plantar flexors and dorsiflexors at both angular velocities compared with guards, as well as significantly (p < 0.05) stronger plantar flexors at 30°·s−1 and dorsiflexors at 60°·s−1 compared with forwards. Our results suggest that the positional differences in quadriceps and hamstring muscle strength of elite male basketball players are the result of respective differences in body size, whereas factors other than body size are responsible for the positional differences in ankle plantar flexor and dorsiflexor strength. These results together with the presented normative isokinetic leg strength data can be useful for coaches and therapists when designing and evaluating position-specific strength training and rehabilitation programs of elite male basketball players.

Almost twenty years after the pioneering efforts of Cardelli, the programming languages community is vigorously pursuing ways to incorporate Fω-style indexed types into programming languages. This paper advocates Concoqtion, a practical approach to adding such highly expressive types to full-fledged programming languages. The approach is applied to MetaOCaml using the Coq proof checker to conservatively extend Hindley-Milner type inference. The implementation of MetaOCaml Concoqtion requires minimal modifications to the syntax, the type checker, and the compiler; and yields a language comparable in notation to the leading proposals. The resulting language provides unlimited expressiveness in the type system while maintaining decidability. Furthermore, programmers can take advantage of a wide range of libraries not only for the programming language but also for the indexed types. Programming in MetaOCaml Concoqtion is illustrated with small examples and a case study implementing a statically-typed domain-specific language.

Building program generators that do not duplicate generated code can be challenging. At the same time, code duplication can easily increase both generation time and runtime of generated programs by an exponential factor. We identify an instance of this problem that can arise when memoized functions are staged. Without addressing this problem, it would be impossible to effectively stage dynamic programming algorithms. Intuitively, direct staging undoes the effect of memoization. To solve this problem once and for all, and for any function that uses memoization, we propose a staged monadic combinator library. Experimental results confirm that the library works as expected. Preliminary results also indicate that the library is useful even when memoization is not used.

This paper addresses the question of how to extend OCaml’s Hindley-Milner type system with types indexed by logical propositions and proofs of the Coq theorem prover, thereby providing an expressive and extensible mechanism for ensuring fine-grained program invariants. We propose adopting the approached used by Shao et al. for certified binaries. This approach maintains a phase distinction between the computational and logical languages, thereby limiting effects and non-termination to the computational language, and maintaining the decidability of the type system. The extension subsumes language features such as impredicative first-class (higher-rank) polymorphism and type operators, that are notoriously difficult to integrate with the Hindley-Milner style of type inference that is used in OCaml. We make the observation that these features can be more easily integrated with type inference if the inference algorithm is free to adapt the order in which it solves typing constraints to each program. To this end we define a novel “order-free” type inference algorithm. The key enabling technology is a graph representation of constraints and a constraint solver that performs Hindley-Milner inference with just three graph rewrite rules.