Semantic Pill 20

Semantic Pill 20



  • Pig and Pig Latin are a Platform and a language to create MapReduce programs within Hadoop;
  • Q-Learning algorithm, it’s a “Reinforced learning” process, where an agent gathers as much as possible positive (rewarding) experience.  The two figures below show how to face an AI, Artificial Intelligence algorithm of Q-learning. The house has 6 “rooms”, five, from 0 to 4 inside the house and sixth outside as an open room. The challenge is to acquire positive experience to train an agent to get out of the house (Finish) beginning on room 2. We may “go” at random starting from “state” 2 and building a “table” or matrix of “rewards”: for example 0 if going “wrong” and 100 if going “right”. We may extend this tiny and basic model to any degree of complexity, trying to find the right way from any place within any labyrinth;
  • Quality of research data, and Quality of research, are sometimes forgotten items. This paper deals with a double polemic theme: quality and how research - in general terms- is really performed. The analysis was extended to three disciplinary domains applied by the European Science Foundation: Physical Sciences and Engineering, Social Sciences and Humanities, and Life Sciences. Finally Social Sciences and Humanities and Life Sciences were summarized as one.


The attitude in Physical Sciences and Engineering would seem to be that quality control of data can best be effectuated through citation of datasets and quality-related comments on those datasets which are made available through Open Access data publications. No need is expressed for codes of conduct, training in data management, or peer review of data that is published together with articles. In Life Sciences there is first and foremost a need for a code of conduct for dealing with data. Training in data management fits in with this. A direct judgment on quality can be given through peer review of the data that is published together with articles and through quality-related comments, a derived judgment through data publications and citations. Open access to data does not score highly. Interestingly enough, Life Sciences are ahead of the other disciplines as regards open access to articles.


Warning: we have to take into account that quality of data is essential and a necessary condition not only for Big Data but for everything: we behave based on data.


  • Finally RTDP has many meanings for example Real Time Data Platform and Real Time Dynamic Programming and within this appears as related to the semantic under study in this pill the theme “Learning to Act using Real Time Dynamic Programming” of which we reproduce here part of its abstract:


Researchers have argued that DP provide an appropriate basis real time control as well as for learning when the system under control is incompletely known. RTDP is DP based algorithm by which an embedded system can improve its performance with experience. It is a generalization of Korf’s Learning Real Time - A* algorithm to problems involving uncertainty;



 Source: Q-learning from