MML, Hybrid Bayesian Network Graphical Models, Statistical Consistency, Invariance and Uniqueness

2015 Eighth International Conference on Advances in Pattern Recognition (ICAPR)

Ray

2015

Self Cite

Abstract-MML (Minimum Message Length) has emerged as a powerful tool in inductive inference of discrete, continuous and hybrid structures. The Probabilistic Finite State Automaton (PFSA) is one such discrete structure that needs to be inferred for classes of problems in the field of Computer Science including artificial intelligence, pattern recognition and data mining. MML has also served as a viable tool in many classes of problems in the field of Machine Learning including both supervised and unsupervised learning. The classification problem is the most common among them. This research is a two-fold solution to a problem where one part focusses on the best inferred PFSA using MML and the second part focusses on the classification problem of Spam Detection. Using the best PFSA inferred in part 1, the Spam Detection theory has been tested using MML on a publicly available Enron Spam dataset. The filter was evaluated on various performance parameters like precision and recall. The evaluation was also done taking into consideration the cost of misclassification in terms of weighted accuracy rate and weighted error rate. The results of our empirical evaluation indicate the classification accuracy to be around 93%, which outperforms well-known established spam filters.

Section: Discussionmentioning

confidence: 99%

MML inference of Finite State Automata for probabilistic spam detection

Saikrishna

2015 Eighth International Conference on Advances in Pattern Recognition (ICAPR)

Ray

2015

Self Cite

“…Indeed, recalling the notion of MML Bayesian network (see, e.g., [7, sec. 7.4 [10]), normalizing and breaking down tables into new tables can be thought of as a (MML) Bayesian net analysis -using the fact that (in some sense) databases could be said to have no noise. And, in similar manner, (the notion of) attribute inheritance (where different types of employee -such as pilot and engineer -have their own specific attributes as well as inheriting common employee attributes) can also be inferred using MML.…”

Section: Discussionmentioning

confidence: 99%

“…Minimum message length (MML), introduced by Wallace and Boulton [4], is an important stream of studying the complexity of a data set [4][5][6][7][8][9][10]. It is based on Shannon's theory of information and equivalently on the theory of Turing machines and Kolmogorov complexity [6,7,10].…”

Section: Minimum Message Lengthmentioning

confidence: 99%

See 1 more Smart Citation

Database Normalization as a By-product of Minimum Message Length Inference

AI 2010: Advances in Artificial Intelligence

Zaidi

2010

Self Cite

Abstract. Database normalization is a central part of database design in which we re-organise the data stored so as to progressively ensure that as few anomalies occur as possible upon insertions, deletions and/or modifications. Successive normalizations of a database to higher normal forms continue to reduce the potential for such anomalies. We show here that database normalization follows as a consequence (or special case, or byproduct) of the Minimum Message Length (MML) principle of machine learning and inductive inference. In other words, someone (previously) oblivious to database normalization but well-versed in MML could examine a database and -using MML considerations alone -normalise it, and even discover the notion of attribute inheritance.

“…In particular, Solomonoff's theory of inductive inference [45], the Minimum Message Length (MML) principle [55,56,54,7], algorithmic information theory [4], Kolmogorov complexity [31,45] and compression theory paved the way in the 1990s for a new approach for defining and measuring intelligence based on algorithmic information theory and two-part compression.…”

Section: Evaluating Cognitive Abilities In the Machine Kingdommentioning

confidence: 99%

On Potential Cognitive Abilities in the Machine Kingdom

Hernández-Orallo

2013

Minds & Machines

Self Cite

Animals, including humans, are usually judged on what they could become, rather than what they are. Many physical and cognitive abilities in the 'animal kingdom' are only acquired (to a given degree) when the subject reaches a certain stage of development, which can be accelerated or spoilt depending on how the environment, training or education is. The term 'potential ability' usually refers to how quick and likely the process of attaining the ability is. In principle, things should not be different for the 'machine kingdom'. While machines can be characterised by a set of cognitive abilities, and measuring them is already a big challenge, known as 'universal psychometrics', a more informative, and yet more challenging, goal would be to also determine the potential cognitive abilities of a machine. In this paper we investigate the notion of potential cognitive ability for machines, focussing especially on universality and intelligence. We consider several machine characterisations (non-interactive and interactive) and give definitions for each case, considering permanent and temporal potentials. From these definitions, we analyse the relation between some potential abilities, we bring out the dependency on the environment distribution and we suggest some ideas about how potential abilities can be measured. Finally, we also analyse the potential of environments at different levels and briefly discuss whether machines should be designed to be intelligent or potentially intelligent.