arXiv subjects

Scott Sanner

Publications and source records attributed to Scott Sanner.

89 records · Page 5Linked to original sources

arXiv2020

ε-BMC: A Bayesian Ensemble Approach to Epsilon-Greedy Exploration in Model-Free Reinforcement Learning

Resolving the exploration-exploitation trade-off remains a fundamental problem in the design and implementation of reinforcement learning (RL) algorithms. In this paper, we focus on model-free RL using the epsilon-greedy exploration policy, which despite its simplicity, remains one of the most frequently used forms of exploration. However, a key limitation of this policy is the specification of $\varepsilon$. In this paper, we provide a novel Bayesian perspective of $\varepsilon$ as a measure of the uniformity of the Q-value function. We introduce a closed-form Bayesian model update based on Bayesian model combination (BMC), based on this new perspective, which allows us to adapt $\varepsilon$ using experiences from the environment in constant time with monotone convergence guarantees. We demonstrate that our proposed algorithm, $\varepsilon$-\texttt{BMC}, efficiently balances exploration and exploitation on different problems, performing comparably or outperforming the best tuned fixed annealing schedules and an alternative data-dependent $\varepsilon$ adaptation scheme proposed in the literature.

cs.LG↗

arXiv2020

Contextual Policy Transfer in Reinforcement Learning Domains via Deep Mixtures-of-Experts

In reinforcement learning, agents that consider the context, or current state, when selecting source policies for transfer have been shown to outperform context-free approaches. However, none of the existing approaches transfer knowledge contextually from model-based learners to a model-free learner. This could be useful, for instance, when source policies are intentionally learned on diverse simulations with plentiful data but transferred to a real-world setting with limited data. In this paper, we assume knowledge of estimated source task dynamics and policies, and common sub-goals but different dynamics. We introduce a novel deep mixture-of-experts formulation for learning state-dependent beliefs over source task dynamics that match the target dynamics using state trajectories collected from the target task. The mixture model is easy to interpret, demonstrates robustness to estimation errors in dynamics, and is compatible with most learning algorithms. We then show how this model can be incorporated into standard policy reuse frameworks, and demonstrate its effectiveness on benchmarks from OpenAI-Gym.

cs.LG↗

arXiv2020

Bayesian Experience Reuse for Learning from Multiple Demonstrators

Learning from demonstrations (LfD) improves the exploration efficiency of a learning agent by incorporating demonstrations from experts. However, demonstration data can often come from multiple experts with conflicting goals, making it difficult to incorporate safely and effectively in online settings. We address this problem in the static and dynamic optimization settings by modelling the uncertainty in source and target task functions using normal-inverse-gamma priors, whose corresponding posteriors are, respectively, learned from demonstrations and target data using Bayesian neural networks with shared features. We use this learned belief to derive a quadratic programming problem whose solution yields a probability distribution over the expert models. Finally, we propose Bayesian Experience Reuse (BERS) to sample demonstrations in accordance with this distribution and reuse them directly in new tasks. We demonstrate the effectiveness of this approach for static optimization of smooth functions, and transfer learning in a high-dimensional supply chain problem with cost uncertainty.

cs.LG↗

arXiv2020

Compact and Efficient Encodings for Planning in Factored State and Action Spaces with Learned Binarized Neural Network Transition Models

In this paper, we leverage the efficiency of Binarized Neural Networks (BNNs) to learn complex state transition models of planning domains with discretized factored state and action spaces. In order to directly exploit this transition structure for planning, we present two novel compilations of the learned factored planning problem with BNNs based on reductions to Weighted Partial Maximum Boolean Satisfiability (FD-SAT-Plan+) as well as Binary Linear Programming (FD-BLP-Plan+). Theoretically, we show that our SAT-based Bi-Directional Neuron Activation Encoding is asymptotically the most compact encoding relative to the current literature and supports Unit Propagation (UP) -- an important property that facilitates efficiency in SAT solvers. Experimentally, we validate the computational efficiency of our Bi-Directional Neuron Activation Encoding in comparison to an existing neuron activation encoding and demonstrate the ability to learn complex transition models with BNNs. We test the runtime efficiency of both FD-SAT-Plan+ and FD-BLP-Plan+ on the learned factored planning problem showing that FD-SAT-Plan+ scales better with increasing BNN size and complexity. Finally, we present a finite-time incremental constraint generation algorithm based on generalized landmark constraints to improve the planning accuracy of our encodings through simulated or real-world interaction.

cs.AI↗

arXiv2020

Optimizing Search API Queries for Twitter Topic Classifiers Using a Maximum Set Coverage Approach

Twitter has grown to become an important platform to access immediate information about major events and dynamic topics. As one example, recent work has shown that classifiers trained to detect topical content on Twitter can generalize well beyond the training data. Since access to Twitter data is hidden behind a limited search API, it is impossible (for most users) to apply these classifiers directly to the Twitter unfiltered data streams ("firehose"). Rather, applications must first decide what content to retrieve through the search API before filtering that content with topical classifiers. Thus, it is critically important to query the Twitter API relative to the intended topical classifier in a way that minimizes the amount of negatively classified data retrieved. In this paper, we propose a sequence of query optimization methods that generalize notions of the maximum coverage problem to find the subset of query terms within the API limits that cover most of the topically relevant tweets without sacrificing precision. We evaluate the proposed methods on a large dataset of Twitter data collected during 2013 and 2014 labeled using manually curated hashtags for eight topics. Among many insights, our analysis shows that the best of the proposed methods can significantly outperform the firehose on precision and F1-score while achieving high recall within strict API limitations.

cs.IR↗

arXiv2019

Reward Potentials for Planning with Learned Neural Network Transition Models

Optimal planning with respect to learned neural network (NN) models in continuous action and state spaces using mixed-integer linear programming (MILP) is a challenging task for branch-and-bound solvers due to the poor linear relaxation of the underlying MILP model. For a given set of features, potential heuristics provide an efficient framework for computing bounds on cost (reward) functions. In this paper, we model the problem of finding optimal potential bounds for learned NN models as a bilevel program, and solve it using a novel finite-time constraint generation algorithm. We then strengthen the linear relaxation of the underlying MILP model by introducing constraints to bound the reward function based on the precomputed reward potentials. Experimentally, we show that our algorithm efficiently computes reward potentials for learned NN models, and that the overhead of computing reward potentials is justified by the overall strengthening of the underlying MILP model for the task of planning over long horizons.

cs.AI↗

arXiv2018

Noise Contrastive Estimation for Scalable Linear Models for One-Class Collaborative Filtering

Previous highly scalable one-class collaborative filtering methods such as Projected Linear Recommendation (PLRec) have advocated using fast randomized SVD to embed items into a latent space, followed by linear regression methods to learn personalized recommendation models per user. Unfortunately, naive SVD embedding methods often exhibit a popularity bias that skews the ability to accurately embed niche items. To address this, we leverage insights from Noise Contrastive Estimation (NCE) to derive a closed-form, efficiently computable "depopularized" embedding. While this method is not ideal for direct recommendation using methods like PureSVD since popularity still plays an important role in recommendation, we find that embedding followed by linear regression to learn personalized user models in a novel method we call NCE-PLRec leverages the improved item embedding of NCE while correcting for its popularity unbiasing in final recommendations. An analysis of the recommendation popularity distribution demonstrates that NCE-PLRec uniformly distributes its recommendations over the popularity spectrum while other methods exhibit distinct biases towards specific popularity subranges, thus artificially restricting their recommendations. Empirically, NCE-PLRec outperforms state-of-the-art methods as well as various ablations of itself on a variety of large-scale recommendation datasets.

cs.IR↗

arXiv2018

Conditional Inference in Pre-trained Variational Autoencoders via Cross-coding

Variational Autoencoders (VAEs) are a popular generative model, but one in which conditional inference can be challenging. If the decomposition into query and evidence variables is fixed, conditional VAEs provide an attractive solution. To support arbitrary queries, one is generally reduced to Markov Chain Monte Carlo sampling methods that can suffer from long mixing times. In this paper, we propose an idea we term cross-coding to approximate the distribution over the latent variables after conditioning on an evidence assignment to some subset of the variables. This allows generating query samples without retraining the full VAE. We experimentally evaluate three variations of cross-coding showing that (i) they can be quickly optimized for different decompositions of evidence and query and (ii) they quantitatively and qualitatively outperform Hamiltonian Monte Carlo.

stat.ML↗

arXiv2018

Aesthetic Features for Personalized Photo Recommendation

Many photography websites such as Flickr, 500px, Unsplash, and Adobe Behance are used by amateur and professional photography enthusiasts. Unlike content-based image search, such users of photography websites are not just looking for photos with certain content, but more generally for photos with a certain photographic "aesthetic". In this context, we explore personalized photo recommendation and propose two aesthetic feature extraction methods based on (i) color space and (ii) deep style transfer embeddings. Using a dataset from 500px, we evaluate how these features can be best leveraged by collaborative filtering methods and show that (ii) provides a significant boost in photo recommendation performance.

cs.IR↗

arXiv2017

Scalable Planning with Tensorflow for Hybrid Nonlinear Domains

Given recent deep learning results that demonstrate the ability to effectively optimize high-dimensional non-convex functions with gradient descent optimization on GPUs, we ask in this paper whether symbolic gradient optimization tools such as Tensorflow can be effective for planning in hybrid (mixed discrete and continuous) nonlinear domains with high dimensional state and action spaces? To this end, we demonstrate that hybrid planning with Tensorflow and RMSProp gradient descent is competitive with mixed integer linear program (MILP) based optimization on piecewise linear planning domains (where we can compute optimal solutions) and substantially outperforms state-of-the-art interior point methods for nonlinear planning domains. Furthermore, we remark that Tensorflow is highly scalable, converging to a strong plan on a large-scale concurrent domain with a total of 576,000 continuous action parameters distributed over a horizon of 96 time steps and 100 parallel instances in only 4 minutes. We provide a number of insights that clarify such strong performance including observations that despite long horizons, RMSProp avoids both the vanishing and exploding gradient problems. Together these results suggest a new frontier for highly scalable planning in nonlinear hybrid domains by leveraging GPUs and the power of recent advances in gradient descent with highly optimized toolkits like Tensorflow.

cs.LG↗

arXiv2017

Expecting to be HIP: Hawkes Intensity Processes for Social Media Popularity

Modeling and predicting the popularity of online content is a significant problem for the practice of information dissemination, advertising, and consumption. Recent work analyzing massive datasets advances our understanding of popularity, but one major gap remains: To precisely quantify the relationship between the popularity of an online item and the external promotions it receives. This work supplies the missing link between exogenous inputs from public social media platforms, such as Twitter, and endogenous responses within the content platform, such as YouTube. We develop a novel mathematical model, the Hawkes intensity process, which can explain the complex popularity history of each video according to its type of content, network of diffusion, and sensitivity to promotion. Our model supplies a prototypical description of videos, called an endo-exo map. This map explains popularity as the result of an extrinsic factor - the amount of promotions from the outside world that the video receives, acting upon two intrinsic factors - sensitivity to promotion, and inherent virality. We use this model to forecast future popularity given promotions on a large 5-months feed of the most-tweeted videos, and found it to lower the average error by 28.6% from approaches based on popularity history. Finally, we can identify videos that have a high potential to become viral, as well as those for which promotions will have hardly any effect.

cs.SI↗

arXiv2017

Stochastic Planning and Lifted Inference

Lifted probabilistic inference (Poole, 2003) and symbolic dynamic programming for lifted stochastic planning (Boutilier et al, 2001) were introduced around the same time as algorithmic efforts to use abstraction in stochastic systems. Over the years, these ideas evolved into two distinct lines of research, each supported by a rich literature. Lifted probabilistic inference focused on efficient arithmetic operations on template-based graphical models under a finite domain assumption while symbolic dynamic programming focused on supporting sequential decision-making in rich quantified logical action models and on open domain reasoning. Given their common motivation but different focal points, both lines of research have yielded highly complementary innovations. In this chapter, we aim to help close the gap between these two research areas by providing an overview of lifted stochastic planning from the perspective of probabilistic inference, showing strong connections to other chapters in this book. This also allows us to define Generalized Lifted Inference as a paradigm that unifies these areas and elucidates open problems for future research that can benefit both lifted inference and stochastic planning.

cs.AI↗

arXiv2016

On the Mathematical Relationship between Expected n-call@k and the Relevance vs. Diversity Trade-off

It has been previously noted that optimization of the n-call@k relevance objective (i.e., a set-based objective that is 1 if at least n documents in a set of k are relevant, otherwise 0) encourages more result set diversification for smaller n, but this statement has never been formally quantified. In this work, we explicitly derive the mathematical relationship between expected n-call@k and the relevance vs. diversity trade-off --- through fortuitous cancellations in the resulting combinatorial optimization, we show the trade-off is a simple and intuitive function of n (notably independent of the result set size k e n), where diversification increases as n approaches 1.

cs.IR↗

arXiv2013

Bounded Approximate Symbolic Dynamic Programming for Hybrid MDPs

Recent advances in symbolic dynamic programming (SDP) combined with the extended algebraic decision diagram (XADD) data structure have provided exact solutions for mixed discrete and continuous (hybrid) MDPs with piecewise linear dynamics and continuous actions. Since XADD-based exact solutions may grow intractably large for many problems, we propose a bounded error compression technique for XADDs that involves the solution of a constrained bilinear saddle point problem. Fortuitously, we show that given the special structure of this problem, it can be expressed as a bilevel linear programming problem and solved to optimality in finite time via constraint generation, despite having an infinite set of constraints. This solution permits the use of efficient linear program solvers for XADD compression and enables a novel class of bounded approximate SDP algorithms for hybrid MDPs that empirically offers order-of-magnitude speedups over the exact solution in exchange for a small approximation error.

cs.AI↗

arXiv2012

Approximate Linear Programming for First-order MDPs

We introduce a new approximate solution technique for first-order Markov decision processes (FOMDPs). Representing the value function linearly w.r.t. a set of first-order basis functions, we compute suitable weights by casting the corresponding optimization as a first-order linear program and show how off-the-shelf theorem prover and LP software can be effectively used. This technique allows one to solve FOMDPs independent of a specific domain instantiation; furthermore, it allows one to determine bounds on approximation error that apply equally to all domain instantiations. We apply this solution technique to the task of elevator scheduling with a rich feature space and multi-criteria additive reward, and demonstrate that it outperforms a number of intuitive, heuristicallyguided policies.

cs.AI↗

arXiv2012

Practical Linear Value-approximation Techniques for First-order MDPs

Recent work on approximate linear programming (ALP) techniques for first-order Markov Decision Processes (FOMDPs) represents the value function linearly w.r.t. a set of first-order basis functions and uses linear programming techniques to determine suitable weights. This approach offers the advantage that it does not require simplification of the first-order value function, and allows one to solve FOMDPs independent of a specific domain instantiation. In this paper, we address several questions to enhance the applicability of this work: (1) Can we extend the first-order ALP framework to approximate policy iteration to address performance deficiencies of previous approaches? (2) Can we automatically generate basis functions and evaluate their impact on value function quality? (3) How can we decompose intractable problems with universally quantified rewards into tractable subproblems? We propose answers to these questions along with a number of novel optimizations and provide a comparative empirical evaluation on logistics problems from the ICAPS 2004 Probabilistic Planning Competition.

cs.AI↗

arXiv2012

Symbolic Dynamic Programming for Discrete and Continuous State MDPs

Many real-world decision-theoretic planning problems can be naturally modeled with discrete and continuous state Markov decision processes (DC-MDPs). While previous work has addressed automated decision-theoretic planning for DCMDPs, optimal solutions have only been defined so far for limited settings, e.g., DC-MDPs having hyper-rectangular piecewise linear value functions. In this work, we extend symbolic dynamic programming (SDP) techniques to provide optimal solutions for a vastly expanded class of DCMDPs. To address the inherent combinatorial aspects of SDP, we introduce the XADD - a continuous variable extension of the algebraic decision diagram (ADD) - that maintains compact representations of the exact value function. Empirically, we demonstrate an implementation of SDP with XADDs on various DC-MDPs, showing the first optimal automated solutions to DCMDPs with linear and nonlinear piecewise partitioned value functions and showing the advantages of constraint-based pruning for XADDs.

cs.AI↗