1,721,030 research outputs found
Reflection mechanisms to combine Prolog databases
No full textBy using practical examples, this paper outlines the power of reflection mechanisms for logic programming systems in the domain of knowledge structuring. In particular, it presents an extension of Prolog, where separate databases can be handled as first‐class objects. Different forms of database combination such as inheritance and dynamic context extension/contraction are specified and implemented in a dynamic and flexible way through reflection. The main aim is to broaden the application area of logic programming to encompass most of the paradigms needed by systems that use artificial intelligence techniques. Practical results presented in the paper show that logic programs that use reflection can be shorter, more readable and efficient than those using more conventional full meta‐interpretation techniques. Full meta‐interpretation, however, is more general than reflection
Learning hierarchical probabilistic logic programs
No full textProbabilistic logic programming (PLP) combines logic programs and probabilities. Due to its expressiveness and simplicity, it has been considered as a powerful tool for learning and reasoning in relational domains characterized by uncertainty. Still, learning the parameter and the structure of general PLP is computationally expensive due to the inference cost. We have recently proposed a restriction of the general PLP language called hierarchical PLP (HPLP) in which clauses and predicates are hierarchically organized. HPLPs can be converted into arithmetic circuits or deep neural networks and inference is much cheaper than for general PLP. In this paper we present algorithms for learning both the parameters and the structure of HPLPs from data. We first present an algorithm, called parameter learning for hierarchical probabilistic logic programs (PHIL) which performs parameter estimation of HPLPs using gradient descent and expectation maximization. We also propose structure learning of hierarchical probabilistic logic programming (SLEAHP), that learns both the structure and the parameters of HPLPs from data. Experiments were performed comparing PHIL and SLEAHP with PLP and Markov Logic Networks state-of-the art systems for parameter and structure learning respectively. PHIL was compared with EMBLEM, ProbLog2 and Tuffy and SLEAHP with SLIPCOVER, PROBFOIL+, MLB-BC, MLN-BT and RDN-B. The experiments on five well known datasets show that our algorithms achieve similar and often better accuracies but in a shorter time
Symbolic DNN-Tuner: A Python and ProbLog-based system for optimizing Deep Neural Networks hyperparameters
No full textThe application of deep learning models to increasingly complex contexts has led to a rise in the complexity of the models themselves. Due to this, there is an increase in the number of hyper-parameters (HPs) to be set and Hyper-Parameter Optimization (HPO) algorithms occupy a fundamental role in deep learning. Bayesian Optimization (BO) is the state-of-the-art of HPO for deep learning models. BO keeps track of past results and uses them to build a probabilistic model, building a probability density of HPs. This work aims to improve BO applied to Deep Neural Networks (DNNs) by an analysis of the results of the network on training and validation sets. This analysis is obtained by applying symbolic tuning rules, implemented in Probabilistic Logic Programming (PLP). The resulting system, called Symbolic DNN-Tuner, logically evaluates the results obtained from the training and the validation phase and, by applying symbolic tuning rules, fixes the network architecture, and its HPs, leading to improved performance. In this paper, we present the general system and its implementation. We also show its graphical interface and a simple example of execution
Automatic Setting of DNN Hyper-Parameters by Mixing Bayesian Optimization and Tuning Rules
No full textDeep learning techniques play an increasingly important role in industrial and research environments due to their outstanding results. However, the large number of hyper-parameters to be set may lead to errors if they are set manually. The state-of-the-art hyper-parameters tuning methods are grid search, random search, and Bayesian Optimization. The first two methods are expensive because they try, respectively, all possible combinations and random combinations of hyper-parameters. Bayesian Optimization, instead, builds a surrogate model of the objective function, quantifies the uncertainty in the surrogate using Gaussian Process Regression and uses an acquisition function to decide where to sample the new set of hyper-parameters. This work faces the field of Hyper-Parameters Optimization (HPO). The aim is to improve Bayesian Optimization applied to Deep Neural Networks. For this goal, we build a new algorithm for evaluating and analyzing the results of the network on the training and validation sets and use a set of tuning rules to add new hyper-parameters and/or to reduce the hyper-parameter search space to select a better combination
GRD-Net: Generative-Reconstructive-Discriminative Anomaly Detection with Region of Interest Attention Module
No full textAnomaly detection is nowadays increasingly used in industrial applications and processes. One of the main fields of the appliance is the visual inspection for surface anomaly detection, which aims to spot regions that deviate from regularity and consequently identify abnormal products. Defect localization is a key task that is usually achieved using a basic comparison between generated image and the original one, implementing some blob analysis or image-editing algorithms in the postprocessing step, which is very biased towards the source dataset, and they are unable to generalize. Furthermore, in industrial applications, the totality of the image is not always interesting but could be one or some regions of interest (ROIs), where only in those areas there are relevant anomalies to be spotted. For these reasons, we propose a new architecture composed by two blocks. The first block is a generative adversarial network (GAN), based on a residual autoencoder (ResAE), to perform reconstruction and denoising processes, while the second block produces image segmentation, spotting defects. This method learns from a dataset composed of good products and generated synthetic defects. The discriminative network is trained using a ROI for each image contained in the training dataset. The network will learn in which area anomalies are relevant. This approach guarantees the reduction of using preprocessing algorithms, formerly developed with blob analysis and image-editing procedures. To test our model, we used challenging MVTec anomaly detection datasets and an industrial large dataset of pharmaceutical BFS strips of vials. This set constitutes a more realistic use case of the aforementioned network
Symbolic DNN-Tuner
No full textHyper-Parameter Optimization (HPO) occupies a fundamental role in Deep Learning systems due to the number of hyper-parameters (HPs) to be set. The state-of-the-art of HPO methods are Grid Search, Random Search and Bayesian Optimization. The first two methods try all possible combinations and random combination of the HPs values, respectively. This is performed in a blind manner, without any information for choosing the new set of HPs values. Bayesian Optimization (BO), instead, keeps track of past results and uses them to build a probabilistic model mapping HPs into a probability density of the objective function. Bayesian Optimization builds a surrogate probabilistic model of the objective function, finds the HPs values that perform best on the surrogate model and updates it with new results. In this paper, we improve BO applied to Deep Neural Network (DNN) by adding an analysis of the results of the network on training and validation sets. This analysis is performed by exploiting rule-based programming, and in particular by using Probabilistic Logic Programming. The resulting system, called Symbolic DNN-Tuner, logically evaluates the results obtained from the training and the validation phase and, by applying symbolic tuning rules, fixes the network architecture, and its HPs, therefore improving performance. We also show the effectiveness of the proposed approach, by an experimental evaluation on literature and real-life datasets
A semantics for Hybrid Probabilistic Logic programs with function symbols
No full textProbabilistic Logic Programming (PLP) is a powerful paradigm for the representation of uncertain relations among objects. Recently, programs with continuous variables, also called hybrid programs, have been proposed and assigned a semantics. Hybrid programs are capable of representing real-world measurements but unfortunately the semantics proposal was imprecise so the definition did not assign a probability to all queries. In this paper, we remedy this and formally define a new semantics for hybrid programs. We prove that the semantics assigns a probability to all queries for a large class of programs
ALIAS: The Abductive LogIc AgentS architecture
No full textALIAS is an architecture whose goal is to introduce abduction in a (logic) multi-agent environment. It implements a distributed protocol to coordinate the reasoning of all the abductive agents in the system, inspired to the basic algorithm for abductive reasoning presented by Kakas and Mancarella. A global knowledge is represented by a set of abduced hypotheses posted in a Linda-like tuple space. A schematic representation of ALIAS architecture is here below.
In ALIAS, agents are equipped with hypothetical reasoning capabilities, obtained by means of abduction. The union of their knowledge bases ends up to generate a global knowledge base, which can change in a dynamic fashion, as a consequence of agent movements. In this framework, agents can perform standard deduction and also abduce new hypotheses, provided that they are consistent with the knowledge of other agents, i.e., the global knowledge represented by the union of the agent logic programs is maintained coherent as it would be if it was owned by a single entity. To this purpose, a mechanism to coordinate agent reasoning is introduced.
Agents in ALIAS are grouped into bunches; each bunch represents an agora where agents can discuss about common arguments and/or cooperate for solving particular problems. ALIAS agents can move from bunch to bunch at runtime (for this reason we call them rambling agents), and collect information (the agent experience) derived from the interactions with other agents in different agoras. More in detail, each agent is characterized by statically defined local knowledge represented by an abductive logic program and possibly by a dynamically built set of assumptions (i.e., its experience). Each bunch is represented by a set of agents and a global knowledge (i.e., the set of hypotheses assumed so far within the bunch) that is dynamically built. While static knowledge is peculiar to each agent and might differ from agent to agent, all agents in the same bunch must agree on the global set of assumed abducibles. To this purpose, a set of integrity constraints is used - together with program clauses - to confirm or discard new hypotheses. Rambling agents can contribute to enlarge the dynamic knowledge of bunches they enter. On the other hand, each rambling agent can improve its experience by moving from bunch to bunch and collecting the set of hypotheses produced in the bunches visited so far.
ALIAS could be usefully employed to implement complex problem solving in distributed systems. For instance, information retrieval and filtering systems could benefit of abductive reasoning to detect inconsistencies in presence of incomplete, multiple and/or conflicting information. Moreover, bunches and Rambling Agents could also be useful in electronic commerce applications since bunches define confined, protected and possibly secure domains, where rambling agents could enter only if they have proper authorizations
An image analysis approach for automatically re-orienteering CT images for dental implants
No full textIn the last decade, computerized tomography (CT) has become the most frequently used imaging modality to obtain a correct pre-operative implant planning. In this work, we present an image analysis and computer vision approach able to identify, from the reconstructed 3D data set, the optimal cutting plane specific to each implant to be planned, in order to obtain the best view of the implant site and to have correct measures. If the patient requires more implants, different cutting planes are automatically identified, and the axial and cross-sectional images can be re-oriented accordingly to each of them. In the paper, we describe the defined algorithms in order to recognize 3D markers (each one aligned with a missed tooth for which an implant has to be planned) in the 3D reconstructed space, and the results in processing red] exams, in terms of effectiveness and precision and reproducibility of the measure
- …
