Reliable prediction of T-cell epitopes using neural networks with
... sparse versus the Blosum sequence-encoding scheme constitutes two different approaches to represent sequence information to the neural network. In the sparse encoding the neural network is given very precise information about the sequence that corresponds to a given training example. One can say tha ...
... sparse versus the Blosum sequence-encoding scheme constitutes two different approaches to represent sequence information to the neural network. In the sparse encoding the neural network is given very precise information about the sequence that corresponds to a given training example. One can say tha ...
A Model Counting Characterization of Diagnoses
... of behavior of a component). Such a constraint in this example would be &' \]"gj `_a_%' \]/" . Diagnosis can become indiscriminate without fault models. It is also easy to see that the consistency-based approach can exploit fault models (when they are specified) to produce more intuitive diag ...
... of behavior of a component). Such a constraint in this example would be &' \]"gj `_a_%' \]/" . Diagnosis can become indiscriminate without fault models. It is also easy to see that the consistency-based approach can exploit fault models (when they are specified) to produce more intuitive diag ...
draft pdf
... fraction of the full space of 2N variable assignments, and the application of Markov’s inequality as in SampleCount’s correctness analysis does not yield interesting upper bounds. Note that systematic model counters like Relsat and Cachet can also be easily extended to provide an upper bound when th ...
... fraction of the full space of 2N variable assignments, and the application of Markov’s inequality as in SampleCount’s correctness analysis does not yield interesting upper bounds. Note that systematic model counters like Relsat and Cachet can also be easily extended to provide an upper bound when th ...
Goal Recognition with Markov Logic Networks for Player
... • Action Type: The type of current action taken by the player, such as moving to a particular location, opening a door, and testing an object using the laboratory’s testing equipment. To avoid data sparsity issues, only the predicate (e.g., OPEN) of the action is considered, ignoring the associated ...
... • Action Type: The type of current action taken by the player, such as moving to a particular location, opening a door, and testing an object using the laboratory’s testing equipment. To avoid data sparsity issues, only the predicate (e.g., OPEN) of the action is considered, ignoring the associated ...
deep variational bayes filters: unsupervised learning of state space
... gradient updates i until reaching 1 after TA annealing iterations. Similar annealing schedules have been applied in, e.g., Ghahramani & Hinton (2000); Mandt et al. (2016); Rezende & Mohamed (2015), where it is shown that they smooth the typically highly non-convex error landscape. Additionally, the ...
... gradient updates i until reaching 1 after TA annealing iterations. Similar annealing schedules have been applied in, e.g., Ghahramani & Hinton (2000); Mandt et al. (2016); Rezende & Mohamed (2015), where it is shown that they smooth the typically highly non-convex error landscape. Additionally, the ...
Semantics and derivation for Stochastic Logic Programs
... a systematic enumeration of proofs for the given goal. The program in this section is more like a standard Prolog interpreter in these ways. The program returns ground atoms from the SLP in descending order of their probability according to either the Normalised or NF interpretation of (note that th ...
... a systematic enumeration of proofs for the given goal. The program in this section is more like a standard Prolog interpreter in these ways. The program returns ground atoms from the SLP in descending order of their probability according to either the Normalised or NF interpretation of (note that th ...
A Probabilistic Extension of the Stable Model
... and Lifschitz 1988) is the language of Answer Set Programming (ASP). Many useful knowledge representation constructs have been introduced in ASP, and several efficient ASP solvers are available. However, like many other logical approaches, ASP is not well suited for handling uncertainty. A Markov Lo ...
... and Lifschitz 1988) is the language of Answer Set Programming (ASP). Many useful knowledge representation constructs have been introduced in ASP, and several efficient ASP solvers are available. However, like many other logical approaches, ASP is not well suited for handling uncertainty. A Markov Lo ...
Basic Probability
... Random variables (RVs) are by convention given capital letters. Say we have the RVs X1 , . . . , Xn . Their values are given using lower case. So for example X1 might be a binary RV taking values true and false, and X2 might be the outcome of rolling a die and therefore taking values one, two, . . . ...
... Random variables (RVs) are by convention given capital letters. Say we have the RVs X1 , . . . , Xn . Their values are given using lower case. So for example X1 might be a binary RV taking values true and false, and X2 might be the outcome of rolling a die and therefore taking values one, two, . . . ...
Estimation of Parameters from Data
... The purpose of process modeling is to predict the behavior of chemical and physical phenomena. More often than not, it is not practical to predict the behaviors involved with ab initio models. Instead, most process models employ phenomenological models to represent the complex underlying physical an ...
... The purpose of process modeling is to predict the behavior of chemical and physical phenomena. More often than not, it is not practical to predict the behaviors involved with ab initio models. Instead, most process models employ phenomenological models to represent the complex underlying physical an ...
PDF
... where K is the maximum number of non-zero entries in a message. In the worst case, the density of states can have an exponential number of non-zero entries (i.e., the finite number of possible energy values, which we will also refer to as “buckets”), for instance when potentials are set to logarithm ...
... where K is the maximum number of non-zero entries in a message. In the worst case, the density of states can have an exponential number of non-zero entries (i.e., the finite number of possible energy values, which we will also refer to as “buckets”), for instance when potentials are set to logarithm ...
Classification of Web Services Using Bayesian Network
... Tabu search. Naive Bayes is special form of Bayesian network that is widely used for classification [3] and clustering [4], but its potential for general probabilistic modeling (i.e., to answer joint, conditional and marginal queries over arbitrary distributions) remains largely unexploited. Naive B ...
... Tabu search. Naive Bayes is special form of Bayesian network that is widely used for classification [3] and clustering [4], but its potential for general probabilistic modeling (i.e., to answer joint, conditional and marginal queries over arbitrary distributions) remains largely unexploited. Naive B ...
Advances in the Understanding and Use of Conditional Independence
... introduction to the literature on the construction of graphical models, the paper by Lauritzen and Jensen can serve as an introduction to the literature on computation with these models, which has evolved into a literature on the general problem of computation in join trees. The basic question invol ...
... introduction to the literature on the construction of graphical models, the paper by Lauritzen and Jensen can serve as an introduction to the literature on computation with these models, which has evolved into a literature on the general problem of computation in join trees. The basic question invol ...
Toward General Analysis of Recursive Probability Models
... language. For expressions that do not contain distributions, the semantics (like the syntax) of the language is the same as that of the normal A-calculus. We have extended this semantics to handle distributions. A distribution may be thought of as a variable whose value will be determined randomly. ...
... language. For expressions that do not contain distributions, the semantics (like the syntax) of the language is the same as that of the normal A-calculus. We have extended this semantics to handle distributions. A distribution may be thought of as a variable whose value will be determined randomly. ...
Aalborg Universitet
... Learning Bayesian network (BN) models from data has been widely studied for the last few years. As a result, two main approaches to learning have been developed: One tests conditional independence constraints, while the other searches the space of models using a score. In this paper we focus on the ...
... Learning Bayesian network (BN) models from data has been widely studied for the last few years. As a result, two main approaches to learning have been developed: One tests conditional independence constraints, while the other searches the space of models using a score. In this paper we focus on the ...
Counting Belief Propagation
... Its underlying idea is rather simple: group together nodes and factors into clusternodes and clusterfeatures that are indistinguishable in terms of messages received and sent given the evidence. Exploiting this symmetry present in the probabilistic model makes it often possible to greatly compress t ...
... Its underlying idea is rather simple: group together nodes and factors into clusternodes and clusterfeatures that are indistinguishable in terms of messages received and sent given the evidence. Exploiting this symmetry present in the probabilistic model makes it often possible to greatly compress t ...
Probabilistic graphical models in artificial intelligence
... For this reason, a new formalism was created, the so-called certainty factors, with the aim of being able to reason with available uncertain rules, without being subject to the requirements of the Theory of Probability. This example was followed by other expert systems and some of these created thei ...
... For this reason, a new formalism was created, the so-called certainty factors, with the aim of being able to reason with available uncertain rules, without being subject to the requirements of the Theory of Probability. This example was followed by other expert systems and some of these created thei ...
Infinite-Horizon Proactive Dynamic DCOPs
... In general, IPD-DCOPs can be solved in an online or offline manner. Online approaches have the benefit of observing the actual values of the random variables during execution and can thus exploit these observations to improve the overall solution quality. However, the downside to online approaches i ...
... In general, IPD-DCOPs can be solved in an online or offline manner. Online approaches have the benefit of observing the actual values of the random variables during execution and can thus exploit these observations to improve the overall solution quality. However, the downside to online approaches i ...
The adversarial stochastic shortest path problem with unknown
... function r, we are given a sequence of reward functions (rt ) describing the rewards at episode t that is assumed to be an individual sequence, that is, no statistical assumption is made about the rewards. More formally, the online loop-free SSP problem M is defined by a tuple (X , A, P, (rt )), whe ...
... function r, we are given a sequence of reward functions (rt ) describing the rewards at episode t that is assumed to be an individual sequence, that is, no statistical assumption is made about the rewards. More formally, the online loop-free SSP problem M is defined by a tuple (X , A, P, (rt )), whe ...
Behaviour Analysis of Multilayer Perceptrons with Multiple Hidden
... input data into a desired response, so they are widely used for pattern classification. With one or two hidden layers, they can approximate virtually any input-output map. They have been shown to approximate the performance of optimal statistical classifiers in difficult problems. Most neural networ ...
... input data into a desired response, so they are widely used for pattern classification. With one or two hidden layers, they can approximate virtually any input-output map. They have been shown to approximate the performance of optimal statistical classifiers in difficult problems. Most neural networ ...
Navigate Like a Cabbie: Probabilistic Reasoning from Observed
... right information and services to users at the appropriate moment. Unlike other methods, which directly model action sequences, PROCAB models the negative utility or cost of each action as a function of contextual variables associated with that action. This allows it to model the reasons for actions ...
... right information and services to users at the appropriate moment. Unlike other methods, which directly model action sequences, PROCAB models the negative utility or cost of each action as a function of contextual variables associated with that action. This allows it to model the reasons for actions ...
Learning with Hierarchical-Deep Models
... models. In particular, we show how we can learn a hierarchical Dirichlet process (HDP) prior over the activities of the top-level features in a DBM, coming to represent both a layered hierarchy of increasingly abstract features and a tree-structured hierarchy of classes. Our model depends minimally ...
... models. In particular, we show how we can learn a hierarchical Dirichlet process (HDP) prior over the activities of the top-level features in a DBM, coming to represent both a layered hierarchy of increasingly abstract features and a tree-structured hierarchy of classes. Our model depends minimally ...
Using Rewards for Belief State Updates in Partially Observable
... Reward signals may provide useful information about the true state of a POMDP system. Although many of the POMDP benchmarks used to evaluate POMDP planning algorithms (e.g., from Tony Cassandra’s repository [3]) have been designed in such a way that rewards do not carry any additional information th ...
... Reward signals may provide useful information about the true state of a POMDP system. Although many of the POMDP benchmarks used to evaluate POMDP planning algorithms (e.g., from Tony Cassandra’s repository [3]) have been designed in such a way that rewards do not carry any additional information th ...
Fast Parameter Learning for Markov Logic Networks Using Bayes Nets
... learning in Markov logic is a convex optimization problem, and thus gradient descent is guaranteed to find the global optimum. However, convergence to this optimum may be extremely slow, partly because the problem is ill-conditioned since different clauses may have very different numbers of satisfyi ...
... learning in Markov logic is a convex optimization problem, and thus gradient descent is guaranteed to find the global optimum. However, convergence to this optimum may be extremely slow, partly because the problem is ill-conditioned since different clauses may have very different numbers of satisfyi ...
Automatic Composition of Music with Methods of Computational
... There are approaches using these methods independent to an optimisation scheme for generation of melodies (see [13, 20] for an overview) what has other needs than an initialisation for optimisation: Using them as a stand-alone method needs them to be more robust and providing nice melodies out of th ...
... There are approaches using these methods independent to an optimisation scheme for generation of melodies (see [13, 20] for an overview) what has other needs than an initialisation for optimisation: Using them as a stand-alone method needs them to be more robust and providing nice melodies out of th ...
Clustering Binary Data with Bernoulli Mixture Models
... arising from the pooling (or mixture) of a finite collection of relatively homogeneous subpopulations. The problem lies, then, in unobserved heterogeneity (Böhning and Seidel, 2003), as it is not known which, or possibly even how many, subpopulations are responsible for producing the data observed. ...
... arising from the pooling (or mixture) of a finite collection of relatively homogeneous subpopulations. The problem lies, then, in unobserved heterogeneity (Böhning and Seidel, 2003), as it is not known which, or possibly even how many, subpopulations are responsible for producing the data observed. ...
Hidden Markov model
A hidden Markov model (HMM) is a statistical Markov model in which the system being modeled is assumed to be a Markov process with unobserved (hidden) states. A HMM can be presented as the simplest dynamic Bayesian network. The mathematics behind the HMM was developed by L. E. Baum and coworkers. It is closely related to an earlier work on the optimal nonlinear filtering problem by Ruslan L. Stratonovich, who was the first to describe the forward-backward procedure.In simpler Markov models (like a Markov chain), the state is directly visible to the observer, and therefore the state transition probabilities are the only parameters. In a hidden Markov model, the state is not directly visible, but output, dependent on the state, is visible. Each state has a probability distribution over the possible output tokens. Therefore the sequence of tokens generated by an HMM gives some information about the sequence of states. Note that the adjective 'hidden' refers to the state sequence through which the model passes, not to the parameters of the model; the model is still referred to as a 'hidden' Markov model even if these parameters are known exactly.Hidden Markov models are especially known for their application in temporal pattern recognition such as speech, handwriting, gesture recognition, part-of-speech tagging, musical score following, partial discharges and bioinformatics.A hidden Markov model can be considered a generalization of a mixture model where the hidden variables (or latent variables), which control the mixture component to be selected for each observation, are related through a Markov process rather than independent of each other. Recently, hidden Markov models have been generalized to pairwise Markov models and triplet Markov models which allow consideration of more complex data structures and the modelling of nonstationary data.