Tommi Jaakkola

We present a novel message passing algorithm for approximating the MAP problem in graphical models. The algorithm is similar in structure to max-product but unlike max-product it always converges,...

We provide a principle for semi-supervised learning based on optimizing the rate of communicating labels for unlabeled points with side information. The side information is expressed in terms of...

Large-Margin Matrix Factorization (2008)

To classify a large number of unlabeled examples we combine a limited number of labeled examples with a Markov random walk representation over the unlabeled examples. The random walk representation...

Nathan Srebro, Jason Rennie, Tommi Jaakkola

We present a novel approach to collaborative prediction, using low-norm instead of low-rank factorizations. The approach is inspired by, and has strong connections to, large-margin linear...

Modeling the combinatorial functions of multiple transcription (2008)

Efficient learning with partially labeled data involves extracting structure from large unlabeled set and combining this information with limited labeled examples. A typical albeit unstated...

Chen-hsiang Yeang, Tommi Jaakkola

factors

Modern classification applications necessitate supplementing the few available labeled examples with unlabeled examples to improve classification performance. We present a new tractable algorithm for...

Algorithmic stability and meta-learning (2008)

Modern classification applications necessitate supplementing the few available labeled examples with unlabeled examples to improve classification performance. We present a new tractable algorithm for...

Andreas Maurer, Tommi Jaakkola

A mechnism of transfer learning is analysed, where samples drawn from different learning tasks of an environment are used to improve the learners performance on a new task. We give a general method...

Fixing Max-Product: Convergent Message Passing Algorithms for MAP LP-Relaxations (2008)

Sparse Matrix Factorization for Analyzing Gene Expression Patterns (2008)

We present a novel message passing algorithm for approximating the MAP problem in graphical models. The algorithm is similar in structure to max-product but unlike max-product it always converges,...

Motivated by the analysis of gene expression data, we develop a new unsupervised modeling technique. Specifically, we study how such data can be modeled via sparse matrix factorization (SMF)....

Satinder P. Singh, Tommi Jaakkola, Michael I. Jordan

It is widely accepted that the use of more compact representations than lookup tables is crucial to scaling reinforcement learning(RL) algorithms to real-world problems. Unfortunately almost all of...

An increasing number of parameter estimation tasks involve the use of at least two information sources, one complete but limited, the other abundant but incomplete. Standard algorithms such as EM (or...

New Outer Bounds on the Marginal Polytope (2008)

We provide a principle for semi-supervised learning based on optimizing the rate of communicating labels for unlabeled points with side information. The side information is expressed in terms of...

David Sontag, Tommi Jaakkola

We give a new class of outer bounds on the marginal polytope, and propose a cutting-plane algorithm for efficiently optimizing over these constraints. When combined with a concave upper bound on the...

Luis Pérez-breva, Chen-hsiang Yeang, Luis E. Ortiz, Tommi Jaakkola

We develop and analyze game-theoretic algorithms for predicting coordinate binding of multiple DNA binding regulators. The allocation of proteins to local neighborhoods and to sites is carried out...

Approximate inference using planar graph (2008)

Approximate inference using planar graph (2008)

decomposition

decomposition

David Sontag, Tommi Jaakkola

We give a new class of outer bounds on the marginal polytope, and propose a cutting-plane algorithm for efficiently optimizing over these constraints. When combined with a concave upper bound on the...

To classify a large number of unlabeled examples we combine a limited number of labeled examples with a Markov random walk representation over the unlabeled examples. The random walk representation...

To classify a large number of unlabeled examples we combine a limited number of labeled examples with a Markov random walk representation over the unlabeled examples. The random walk representation...

Classification with partially labeled data requires using a large number of unlabeled examples (or an estimated marginal P (x)), to further constrain the conditional P (y|x) beyond a few available...

Luis Pérez-breva, Chen-hsiang Yeang, Luis E. Ortiz, Tommi Jaakkola

We develop and analyze game-theoretic algorithms for predicting coordinate binding of multiple DNA binding regulators. The allocation of proteins to local neighborhoods and to sites is carried out...

Journal of Machine Learning Research 7 (2006) 625--644 Submitted 1/05; Revised 9/05; Published 4/06 Stochastic Complexities of Gaussian Mixtures (2008)

In Variational Bayesian, Kazuho Watanabe, I Lab, Tommi Jaakkola

Bayesian learning has been widely used and proved to be effective in many data modeling problems.

Distributed Information Regularization on Graphs (2008)

We provide a principle for semi-supervised learning based on optimizing the rate of communicating labels for unlabeled points with side information.

-- Bayesian networks (2007)

We present a general framework for discriminative estimation based on the maximum entropy principle and its extensions. All calculations involve distributions over structures and/or parameters rather...

Michael I. Jordan, Zoubin Ghahramani, Tommi Jaakkola, Marina Meila, Lawrence Saul

ffl A wide variety of models are being studied in the machine learning literature, including:-- neural networks-- decision trees-- variants of mixture models-- variants of hidden Markov models--...

Tree consistency and bounds on the performance of (2007)

the max-product algorithm and its generalizations

Generalized Low-Rank Approximations (2007)

Tree-based reparameterization framework for analysis of belief propagation and related algorithms

We study the frequent problem of approximating a target matrix with a matrix of lower rank. We provide a simple and efficient (EM) algorithm for solving weighted low rank approximation problems,...

th (2007)

MAP estimation via agreement on (hyper)trees: Message-passing and linear programming approaches

Stable MixeN of Complete and Incomplete Information (2007)

We present a general framework for discriminative estimation based on the maximum entropy principle and its extensions. All calculations involve distributions over structures and/or parameters rather...

massachusac 8 institu2

Stable Mixing of Complete and lncomplete lnformation (2007)

Eective discrimination is essential in many application areas including speech recognition, image classi cation or identication of molecular binding sites in genomic DNA. Statistical approaches used...

Tommi Jaakkola, Hava Siegelmann

An increasing number of parameter estimation tasks involve the use of at least two information sources, one complete but limited, the other abundant but incomplete. Standard algorithms such as EM (or...

MIT LIDS (2007)

In classical large information retrieval systems, the system responds to a user initiated query with a list of results ranked by relevance. The users may further rene their query as needed. This...

Submitted to Neural Computation. On the Convergence of Stochastic Iterative Dynamic Programming Algorithms (2007)

Tommi Jaakkola, Michael I. Jordan, Satinder P. Singh

1 Recent developments in the area of reinforcement learning have yielded a number of new algorithms for the prediction and control of Markovian environments. These algorithms, including the TD()...

Information Regularization with Partially (2007)

Labeled Data Martin, Martin Szummer, Tommi Jaakkola

Classification with partially labeled data requires using a large number of unlabeled examples (or an estimated marginal P (x)), to further constrain the conditional P (y|x) beyond a few available...

Generalized Low-Rank Approximations (2007)

Nathan Srebro Tommi, Tommi Jaakkola

We study the frequent problem of approximating a target matrix with a matrix of lower rank. We provide a simple and efficient (EM) algorithm for solving weighted low rank approximation problems,...

with Naive Bayes (2007)

Automated discovery of functional generality of human gene expression programs (2007)

There are numerous text documents available in electronic form. More and more are becoming available every day. Such documents represent a massive amount of information that is easily accessible....

Georg Kurt Gerber, Robin D Dowell, Tommi Jaakkola, David Gifford

An important research problem in computational biology is the identification of expression programs, sets of co-expressed genes orchestrating normal or pathological processes, and the...

Certified by........................................................................ (2007)

Tommi Jaakkola, Arthur C. Smith

This thesis focuses on the problem of extracting information from informal communication. Textual informal communication, such as e-mail, bulletin boards and blogs, has become a vast information...

Extracting Information from Informal Communication by (2007)

New outer bounds on the marginal polytope (2007)

This thesis focuses on the problem of extracting information from informal communication. Textual informal communication, such as e-mail, bulletin boards and blogs, has become a vast information...

David Sontag, Tommi Jaakkola

We give a new class of outer bounds on the marginal polytope, and propose a cutting-plane algorithm for efficiently optimizing over these constraints. When combined with a concave upper bound on the...

DNA Binding and Games (2006)

Perez-Breva, Luis, Ortiz, Luis E., Yeang, Chen-Hsiang, Jaakkola, Tommi

We propose a game-theoretic approach tolearn and predict coordinate binding of multiple DNA bindingregulators. The framework implements resource constrainedallocation of proteins to local...

DNA Binding and Games (2006)

Perez-Breva, Luis, Ortiz, Luis E., Yeang, Chen-Hsiang, Jaakkola, Tommi

We propose a game-theoretic approach tolearn and predict coordinate binding of multiple DNA bindingregulators. The framework implements resource constrainedallocation of proteins to local...

A Comprehensive Approach to Fusion for Microsensor Networks: Distributed and Hierarchical Inference, Communication, and Adaption (2006)

Fisher III., John W., Cetin, Mujdat, Jaakkola, Tommi, Tsitsiklis, John, Verdu, Sergio, Kulkarni, Sanjeev, ...

This final report summarizes the research and activities under the ODDR&E MURI on Data Fusion in Large Arrays of Microsensors. The report reviews the intellectual themes and research concentration...

On the Convergence of Stochastic Iterative Dynamic Programming Algorithms (2006)

Jaakkola, Tommi, Jordan, Michael I., Singh, Satinder P.

Recent developments in the area of reinforcement learning have yielded a number of new algorithms for the prediction and control of Markovian environments. These algorithms, including the TD lambda)...

Journal of Machine Learning Research 7 (2006) 167--189 Submitted 3/04; Revised 9/05; Published 2/06 MinReg: A Scalable Algorithm for Learning (2006)

Parsimonious Regulatory Networks, Amos Tanay, Tommi Jaakkola

In recent years, there has been a growing interest in applying Bayesian networks and their extensions to reconstruct regulatory networks from gene expression data. Since the gene expression domain...

Mixed membership stochastic block models for relational data with application to protein-protein interactions (2006)

Edoardo M. Airoldi, David M. Blei, Stephen E. Fienberg, Eric P. Xing, Tommi Jaakkola

We develop a model for examining data that consists of pairwise measurements, for example, presence or absence of links between pairs of objects. Examples include protein interactions and gene...

Online Learning of Non-stationary Sequences (2005)

Monteleoni, Claire, Jaakkola, Tommi

We consider an online learning scenario in which the learner can make predictions on the basis of a fixed set of experts. We derive upper and lower relative loss bounds for a class of universal...

Online Learning of Non-stationary Sequences (2005)

Monteleoni, Claire, Jaakkola, Tommi

We consider an online learning scenario in which the learner can make predictions on the basis of a fixed set of experts. We derive upper and lower relative loss bounds for a class of universal...

Validation and refinement of gene-regulatory pathways on a network of physical interactions (2005)

Yeang, Chen-Hsiang, Mak, H Craig, McCuine, Scott, Workman, Christopher, Jaakkola, Tommi, Ideker, Trey

Abstract As genome-scale measurements lead to increasingly complex models of gene regulation, systematic approaches are needed to validate and refine these models. Towards this goal, we describe an...

Generalization error bounds for collaborative prediction with low-rank matrices (2005)

MAP estimation via agreement on (hyper)trees: Message-passing and linear-programming approaches (2005)

We prove generalization error bounds for predicting entries in a partially observed matrix by approximating the observed entries with a low-rank matrix. To do so, we bound the number of sign...

Variational message passing (2005)

We develop an approach for computing provably exact maximum a posteriori (MAP) configurations for a subclass of problems on graphs with cycles. By decomposing the original problem into a convex...

John Winn, Christopher M. Bishop, Tommi Jaakkola

This paper presents Variational Message Passing (VMP), a general purpose algorithm for applying variational inference to a Bayesian Network. Like belief propagation, Variational Message Passing...

Learning module networks (2005)

Eran Segal, Daphne Koller, Nir Friedman, Tommi Jaakkola

Methods for learning Bayesian networks can discover dependency structure between observed variables. Although these methods are useful in many applications, they run into computational and...

Diffusion Kernels on Statistical Manifolds (2005)

John Lafferty, Guy Lebanon, Tommi Jaakkola

A family of kernels for statistical learning is introduced that exploits the geometric structure of statistical models. The kernels are based on the heat equation on the Riemannian manifold defined...

Efficient computation of gapped substring kernels on large alphabets (2005)

Juho Rousu, Tommi Jaakkola

We present a sparse dynamic programming algorithm that, given two strings s and t, a gap penalty λ, and an integer p, computes the value of the gap-weighted length-p subsequences kernel. The...

2005 Yeang et Volume al. 6, Issue 7, Article R62 Open Access Method (2005)

Chen-hsiang Yeang, H Craig Mak, Scott Mccuine, Christopher Workman, Tommi Jaakkola, Trey Ideker

Validation and refinement of gene-regulatory pathways on a network of physical interactions

Learning module networks (2005)

Eran Segal, Daphne Koller, Nir Friedman, Tommi Jaakkola

Methods for learning Bayesian networks can discover dependency structure between observed variables. Although these methods are useful in many applications, they run into computational and...

Variational message passing (2005)

John Winn, Christopher M. Bishop, Tommi Jaakkola

Bayesian inference is now widely established as one of the principal foundations for machine learning. In practice, exact inference is rarely possible, and so a variety of approximation techniques...

Variational message passing (2005)

John Winn, Christopher M. Bishop, Tommi Jaakkola

Bayesian inference is now widely established as one of the principal foundations for machine learning. In practice, exact inference is rarely possible, and so a variety of approximation techniques...

Managing the 802.11 Energy/Performance Tradeoff with Machine Learning (2004)

Monteleoni, Claire, Balakrishnan, Hari, Feamster, Nick, Jaakkola, Tommi

This paper addresses the problem of managing the tradeoff betweenenergy consumption and performance in wireless devices implementingthe IEEE 802.11 standard. To save energy, the 802.11...

Managing the 802.11 Energy/Performance Tradeoff with Machine Learning (2004)

Monteleoni, Claire, Balakrishnan, Hari, Feamster, Nick, Jaakkola, Tommi

This paper addresses the problem of managing the tradeoff betweenenergy consumption and performance in wireless devices implementingthe IEEE 802.11 standard. To save energy, the 802.11...

Large-Margin Matrix Factorization (2004)

Nathan Srebro, Jason Rennie, Tommi Jaakkola

We present a novel approach to collaborative prediction, using low-norm instead of low-rank factorizations. The approach is inspired by, and has strong connections to, large-margin linear...

Physical network models (2004)

Chen-hsiang Yeang, Trey Ideker, Tommi Jaakkola

We develop a new framework for inferring models of transcriptional regulation. The models, which we call physical network models, are annotated molecular interaction graphs. The attributes in the...

Managing the 802.11 energy/performance tradeoff with machine learning (2004)

Claire Monteleoni, Hari Balakrishnan, Nick Feamster, Tommi Jaakkola

Abstract—This paper addresses the problem of managing the tradeoff between energy consumption and performance in wireless devices implementing the IEEE 802.11 standard [1]. To save energy, the...

Generalized Low-Rank Approximations (2003)

Srebro, Nathan, Jaakkola, Tommi

We study the frequent problem of approximating a target matrix with a matrix of lower rank. We provide a simple and efficient (EM) algorithm for solving {\\em weighted} low rank approximation...

Generalized Low-Rank Approximations (2003)

Srebro, Nathan, Jaakkola, Tommi

We study the frequent problem of approximating a target matrix with a matrix of lower rank. We provide a simple and efficient (EM) algorithm for solving {\\em weighted} low rank approximation...

Linear dependent dimensionality reduction (2003)

Information regularization with partially labeled data (2003)

We formulate linear dimensionality reduction as a semi-parametric estimation problem, enabling us to study its asymptotic behavior. We generalize the problem beyond additive Gaussian noise to...

On information regularization (2003)

Classification with partially labeled data requires using a large number of unlabeled examples (or an estimated marginal P (x)), to further constrain the conditional P (y|x) beyond a few available...

Weighted Low-Rank Approximations (2003)

We formulate a principle for classification with the knowledge of the marginal distribution over the data points (unlabeled data). The principle is cast in terms of Tikhonov style regularization...

Nathan Srebro Nati, Tommi Jaakkola

We study the common problem of approximating a target matrix with a matrix of lower rank. We provide a simple and e#cient (EM) algorithm for solving weighted low-rank approximation problems, which,...

Online Learning of Non-stationary Sequences (2003)

Claire Monteleoni, Tommi Jaakkola

We consider an online learning scenario in which the learner can make predictions on the basis of a fixed set of experts. We derive upper and lower relative loss bounds for a class of universal...

Physical Network Models and Multi-source Data Integration (2003)

Chen-hsiang Yeang, Tommi Jaakkola

We develop a new framework for inferring models of transcriptional regulation. The models in this approach, which we call physical models, are constructed on the basis of verifiable molecular...

Linear dependent dimensionality reduction (2003)

Online learning of non-stationary sequences (2003)

We formulate linear dimensionality reduction as a semi-parametric estimation problem, enabling us to study its asymptotic behavior. We generalize the problem beyond additive Gaussian noise to...

Claire Monteleoni, Tommi Jaakkola

We consider an online learning scenario in which the learner can make predictions on the basis of a fixed set of experts. We derive upper and lower relative loss bounds for a class of universal...

Online learning of non-stationary sequences (2003)

Claire Monteleoni, Claire Monteleoni, Tommi Jaakkola, Tommi Jaakkola

We consider an online learning scenario in which the learner can make predictions on the basis of a fixed set of experts. We derive upper and lower relative loss bounds for a class of universal...

Linear Dependent Dimensionality Reduction (2003)

Online Learning of Non-stationary Sequences (2003)

We formulate linear dimensionality reduction as a semi-parametric estimation problem, enabling us to study its asymptotic behavior. We generalize the problem beyond additive Gaussian noise to...

Claire Monteleoni And, Claire Monteleoni, Tommi Jaakkola

We consider an online learning scenario in which the learner can make predictions on the basis of a fixed set of experts. We derive upper and lower relative loss bounds for a class of universal...

Abstract (2002)

Chen-hsiang Yeang, Tommi Jaakkola

We develop a new framework for inferring models of transcriptional regulation. The models in this approach, which we call physical models, are constructed on the basis of verifiable molecular...

C ○ 2004 Kluwer Academic Publishers. Manufactured in The Netherlands. (2002)

Continuation methods for mixing heterogeneous sources (2002)

Tree consistency and bounds on the performance of the max-product algorithm and its generalizations

Automatic Feature Induction for Text Classification (2002)

A number of modern learning tasks involve estimation from heterogeneous information sources. This includes classification with labeled and unlabeled data as well as other problems with analogous...

Tree Consistency and Bounds on the Performance of the Max-Product Algorithm and Its Generalizations (2002)

The Problem: All classifiers require a set of features that can be used to distinguish between different examples. In some cases, such as determining whether a chess position is a winning position,...

Learning from Partially Labeled Data (2002)

Finding the maximum a posteriori (MAP) assignment of a discrete-state distribution speci ed by a graphical model requires solving an integer program. The max-product algorithm, also known as the...

Martin Szummer, Tommi Jaakkola, Tomaso Poggio

The Problem: Learning from data with both labeled training points (x,y pairs) and unlabeled training points (x alone). For the labeled points, supervised learning techniques apply, but they cannot...

Stable Mixing of Complete and Incomplete Information (2001)

Corduneanu, Adrian, Jaakkola, Tommi

An increasing number of parameter estimation tasks involve the use of at least two information sources, one complete but limited, the other abundant but incomplete. Standard algorithms such as EM (or...

Stable Mixing of Complete and Incomplete Information (2001)

Corduneanu, Adrian, Jaakkola, Tommi

An increasing number of parameter estimation tasks involve the use of at least two information sources, one complete but limited, the other abundant but incomplete. Standard algorithms such as EM (or...

Partially labeled classification with markov random walks (2001)

Tree-Based Reparameterization for Approximate Estimation on Loopy graphs (2001)

walks

Martin J. Wainwright, Tommi Jaakkola, Alan S. Willsky

We present a tree-based reparameterization framework that provides a new conceptual view of a large class of iterative algorithms for computing approximate marginals in graphs with cycles. It...

Partially labeled classification with markov random walks (2001)

Improving multi-class text classification with naive bayes (2001)

walks

Active Information Retrieval (2001)

There are numerous text documents available in electronic form. More and more are becoming available every day. Such documents represent a massive amount of information that is easily accessible....

Tommi Jaakkola Mit, Tommi Jaakkola, Hava Siegelmann

In classical large information retrieval systems, the system responds to a user initiated query with a list of results ranked by relevance.

Tree-Based Reparameterization Framework for Approximate Estimation of Stochastic Processes on Graphs With Cycles (2001)

Professional and research experience: (2000)

We present a tree-based reparameterization framework for the approximate estimation of stochastic processes on graphs with cycles. This framework provides a new conceptual view of a large class of...

Nathan Srebro, Supervisor Prof, Sam Roweis, Supervisor Prof, Tommi Jaakkola, ...

I conducted research on alternatively spliced exons using the LEADS expressed sequence database.

Convergence results for single-step on-policy reinforcement-learning algorithms (2000)

Tommi Jaakkola, Michael L. Littman, Csaba Szepesv Ari

Abstract. An important application of reinforcement learning (RL) is to finite-state control problems and one of the most difficult problems in learning for control is balancing the exploration...

Feature selection and dualities in maximum entropy discrimination (2000)

Tony Jebara, Tommi Jaakkola

We present the maximum entropy discrimination (MED) formalism as a regularization approach with information theoretic penalties. By extending discriminative and large margin concepts to a...

Kernel expansions with unlabeled examples (2000)

Convergence results for single-step on-policy reinforcement-learning algorithms (2000)

Modern classification applications necessitate supplementing the few available labeled examples with unlabeled examples to improve classification performance. We present a new tractable algorithm for...

Tommi Jaakkola, Michael L. Littman, Csaba Szepesv Ari

Abstract. An important application of reinforcement learning (RL) is to finite-state control problems and one of the most difficult problems in learning for control is balancing the exploration...

Kernel expansions with unlabeled examples (2000)

Feature selection and dualities in maximum entropy discrimination (2000)

Modern classification applications necessitate supplementing the few available labeled examples with unlabeled examples to improve classification performance. We present a new tractable algorithm for...

Tony Jebara, Tommi Jaakkola

Incorporating feature selection into a classification or regression method often carries a number of advantages. In this paper we formalize feature selection specifically from a discriminative...

Tractable bayesian learning of tree belief networks (2000)

Marina Meila, Tommi Jaakkola

award number DMS--9873442.

Tractable Bayesian Learning of Tree Belief Networks (2000)

Marina Meila, Tommi Jaakkola

In this paper we present decomposable priors, a family of priors over structure and parameters of tree belief nets for which Bayesian learning with complete observations is tractable, in the sense...

Feature Selection and Dualities in Maximum Entropy Discrimination (2000)

Tony Jebara, Tommi Jaakkola

Incorporating feature selection into a classification or regression method often carries a number of advantages. In this paper we formalize feature selection specifically from a discriminative...

Feature Selection and Dualities in Maximum Entropy Discrimination (2000)

Tony Jebara Mit, Tony Jebara, Tommi Jaakkola

Incorporating feature selection into a classification or regression method often carries a number of advantages. In this paper we formalize feature selection specifically from a discriminative...

Tractable bayesian learning of tree belief networks (2000)

Marina Meila, Tommi Jaakkola

In this paper we present decomposable priors, a family of priors over structure and parameters of tree belief nets for which Bayesian learning with complete observations is tractable, in the sense...

Tractable bayesian learning of tree belief networks (2000)

Marina Meilă, Tommi Jaakkola

In this paper we present decomposable priors, a family of priors over structure and parameters of tree belief nets for which Bayesian learning with complete observations is tractable, in the sense...

Jaakkola, Tommi, Meila, Marina, Jebara, Tony

We present a general framework for discriminative estimation based on the maximum entropy principle and its extensions. All calculations involve distributions over structures and/or parameters rather...

Jaakkola, Tommi, Meila, Marina, Jebara, Tony

We present a general framework for discriminative estimation based on the maximum entropy principle and its extensions. All calculations involve distributions over structures and/or parameters rather...

Using the fisher kernel method to detect remote protein homologies (1999)

Tommi Jaakkola, Mark Diekhans, David Haussler

A new method, called the Fisher kernel method, for detecting remote protein homologies is introduced and shown to perform well in classifying protein domains by SCOP superfamily. The method is a...

A Discriminative Framework for Detecting Remote Protein Homologies (1999)

We present a general framework for discriminative estimation based on the maximum entropy principle and its extensions. All calculations involve distributions over structures and/or parameters rather...

Tommi Jaakkola, Mark Diekhans, David Haussler

A new method for detecting remote protein homologies is introduced and shown to perform well in classifying protein domains by SCOP superfamily. The method is a variant of support vector machines...

Variational probabilistic inference and the QMR-DT database (1999)

We present a general framework for discriminative estimation based on the maximum entropy principle and its extensions. All calculations involve distributions over structures and/or parameters rather...

Tommi Jaakkola, Michael I. Jordan

We describe a variational approximation method for efficient inference in large-scale probabilistic models. Variational methods are deterministic procedures that provide approximations to marginal...

y (1999)

Dissertation: Statistical Learning for Intelligent Network Architecture (1999)

z

Karen Sollins, David Clark, Tommi Jaakkola

Broad interest in network architecture, communications, systems engineering, measurement and artificial intelligence. Focus on discovering and exploiting non-utilized information in large complex...

Exploiting Generative Models in Discriminative Classifiers (1998)

Tommi Jaakkola, David Haussler

Generative probability models such as hidden Markov models provide a principled way of treating missing information and dealing with variable length sequences. On the other hand, discriminative...

A Discriminative Framework for Detecting Remote Protein Homologies (1998)

Tommi Jaakkola, Mark Diekhans, David Haussler

A new method for detecting remote protein homologies is introduced and shown to perform well in classifying protein domains by SCOP superfamily. The method is a variant of support vector machines...

Approximating Posterior Distributions in Belief Networks using Mixtures (1998)

Christopher M. Bishop, Neil D. Lawrence, Tommi Jaakkola, Michael I. Jordan

Exact inference in densely connected Bayesian networks is computationally intractable, and so there is considerable interest in developing effective approximation schemes. One approach which has been...

Mean Field Theory for Sigmoid Belief Networks (1996)

Saul, Lawrence K., Jaakkola, Tommi, Jordan, Michael I.

We develop a mean field theory for sigmoid belief networks based on ideas from statistical mechanics. Our mean field theory provides a tractable approximation to the true probability distribution in...

Mean Field Theory for Sigmoid Belief Networks (1996)

Saul, Lawrence K., Jaakkola, Tommi, Jordan, Michael I.

We develop a mean field theory for sigmoid belief networks based on ideas from statistical mechanics. Our mean field theory provides a tractable approximation to the true probability distribution in...

Mean field theory for sigmoid belief networks (1996)

Lawrence K. Saul, Tommi Jaakkola, Michael I. Jordan

This publication can be retrieved by anonymous ftp to publications.ai.mit.edu. We develop a mean field theory for sigmoid belief networks based on ideas from statistical mechanics. Our mean field...

Fast learning by bounding likelihoods in sigmoid type belief networks (1996)

Tommi Jaakkola, Lawrence K. Saul, Michael I. Jordan

Sigmoid type belief networks, a class of probabilistic neural networks, provide a natural framework for compactly representing probabilistic information in a variety of unsupervised and supervised...

Mean field theory for sigmoid belief networks (1996)

Lawrence K. Saul, Tommi Jaakkola, Michael I. Jordan

We develop a mean field theory for sigmoid belief networks based on ideas from statistical mechanics. Our mean field theory provides a tractable approximation to the true probability distribution in...

Computing Upper and Lower Bounds on Likelihoods in Intractable Networks (1996)

Tommi Jaakkola, Michael I. Jordan

We present techniques for computing upper and lower bounds on the likelihoods of partial instantiations of variables in sigmoid and noisy-OR networks. The bounds determine confidence intervals for...

Mean Field Theory for Sigmoid Belief Networks (1996)

Lawrence Saul, Tommi Jaakkola, Michael I. Jordan

In this paper we show how to calculate a rigorous lower bound on the likelihood of observed activities in sigmoid belief networks. We view these networks in the framework of statistical mechanics and...

Mean Field Theory for Sigmoid Belief Networks (1996)

Lawrence K. Saul, Tommi Jaakkola, Michael I. Jordan

We develop a mean field theory for sigmoid belief networks based on ideas from statistical mechanics. Our mean field theory provides a tractable approximation to the true probability distribution in...

Mean Field Theory for Sigmoid Belief Networks (1996)

Lawrence K. Saul, Tommi Jaakkola, Michael I. Jordan

We develop a mean field theory for sigmoid belief networks based on ideas from statistical mechanics. Our mean field theory provides a tractable approximation to the true probability distribution in...

Mean field theory for sigmoid belief networks (1996)

Lawrence K. Saul, Tommi Jaakkola, Michael I. Jordan

We develop a mean eld theory for sigmoid belief networks based on ideas from statistical mechanics. Our mean eld theory provides a tractable approximation to the true probability distribution in...

Reinforcement Learning with Soft State Aggregation (1995)

Satinder P. Singh, Tommi Jaakkola, Michael I. Jordan

It is widely accepted that the use of more compact representations than lookup tables is crucial to scaling reinforcement learning (RL) algorithms to real-world problems. Unfortunately almost all of...

Reinforcement Learning Algorithm for Partially Observable Markov Decision Problems (1995)

Tommi Jaakkola, Satinder P. Singh, Michael I. Jordan

Increasing attention has been paid to reinforcement learning algorithms in recent years, partly due to successes in the theoretical analysis of their behavior in Markov environments. If the Markov...

Reinforcement Learning with Soft State Aggregation (1995)

Satinder Singh Singh, Tommi Jaakkola, Michael I. Jordan

It is widely accepted that the use of more compact representations than lookup tables is crucial to scaling reinforcement learning (RL) algorithms to real-world problems. Unfortunately almost all of...

Reinforcement learning algorithm for partially observable markov decision problems (1995)

Tommi Jaakkola, Satinder P. Singh, Michael I. Jordan

Increasing attention has been paid to reinforcement learning algorithms in recent years, partly due to successes in the theoretical analysis of their behaviorinMarkov environments. If the Markov...

On the convergence of stochastic iterative dynamic programming algorithms (1994)

Tommi Jaakkola, Michael I. Jordan, Satinder P. Singh

Increasing attention has recently been paid to algorithms based on dynamic programming (DP) due to the suitability of DP for learning problems involving control. In stochastic environments where the...

On the convergence of stochastic iterative dynamic programming algorithms (1994)

Tommi Jaakkola, Michael I. Jordan, Satinder P. Singh

Recent developments in the area of reinforcement learning have yielded a number of new algorithms for the prediction and control of Markovian environments. These algorithms, including the TD()...

On the Convergence of Stochastic Iterative Dynamic Programming Algorithms (1994)

Tommi Jaakkola, Michael I. Jordan, Satinder P. Singh

Recent developments in the area of reinforcement learning have yielded a number of new algorithms for the prediction and control of Markovian environments. These algorithms, including the TD()...

On the Convergence of Stochastic Iterative Dynamic Programming Algorithms (1994)

Tommi Jaakkola, Michael I. Jordan, Satinder P. Singh

Recent developments in the area of reinforcement learning have yielded a number of new algorithms for the prediction and control of Markovian environments. These algorithms, including the TD()...

Learning Without State-Estimation in Partially Observable Markovian Decision Processes (1994)

Satinder P. Singh, Tommi Jaakkola, Michael I. Jordan

Reinforcement learning (RL) algorithms provide a sound theoretical basis for building learning control architectures for embedded agents. Unfortunately all of the theory and much of the practice (see...

On the convergence of stochastic iterative dynamic programming algorithms (1994)

Tommi Jaakkola, Michael I. Jordan, Satinder P. Singh

putational Learning at MIT, including funds provided by DARPA under the

On the Convergence of Stochastic Iterative Dynamic Programming Algorithms (1993)

Jaakkola, Tommi, Jordan, Michael I., Singh, Satinder P.

Recent developments in the area of reinforcement learning have yielded a number of new algorithms for the prediction and control of Markovian environments. These algorithms, including the TD(lambda)...

On the Convergence of Stochastic Iterative Dynamic Programming Algorithms (1993)

Jaakkola, Tommi, Jordan, Michael I., Singh, Satinder P.

Recent developments in the area of reinforcement learning have yielded a number of new algorithms for the prediction and control of Markovian environments. These algorithms, including the TD(lambda)...

Validation and refinement of gene-regulatory pathways on a network of physical interactions

Yeang, Chen-Hsiang, Mak, H Craig, McCuine, Scott, Workman, Christopher, Jaakkola, Tommi, Ideker, Trey

A new automated procedure for prioritizing genetic perturbations was used to evaluate 38 candidate regulatory networks in yeast. Further analysis of four high-priority gene knockout experiments...

Validation and refinement of gene-regulatory pathways on a network of physical interactions

Yeang, Chen-Hsiang, Mak, H Craig, McCuine, Scott, Workman, Christopher, Jaakkola, Tommi, Ideker, Trey

A new automated procedure for prioritizing genetic perturbations was used to evaluate 38 candidate regulatory networks in yeast. Further analysis of four high-priority gene knockout experiments...

Recursive Algorithms for Approximating Probabilities in Graphical Models

Tommi Jaakkola, Michael I. Jordan

We develop a recursive node-elimination formalism for efficiently approximating large probabilistic networks. No constraints are set on the network topologies. Yet the formalism can be...

Kernel Expansions with Unlabeled Examples