Amir-massoud Farahmand

Claas Voelcker, Victor Liao, Animesh Garg et al. · gatech, nvidia

Model-based reinforcement learning (MBRL) is a sample efficient technique to obtain control policies, yet unavoidable modeling errors often lead performance deterioration. The model in MBRL is often solely fitted to reconstruct dynamics, state observations in particular, while the impact of model error on the policy is not captured by the training objective. This leads to a mismatch between the intended goal of MBRL, enabling good policy and value learning, and the target of the loss function employed in practice, future state prediction. Naive intuition would suggest that value-aware model learning would fix this problem and, indeed, several solutions to this objective mismatch problem have been proposed based on theoretical analysis. However, they tend to be inferior in practice to commonly used maximum likelihood (MLE) based approaches. In this paper we propose the Value-gradient weighted Model Learning (VaGraM), a novel method for value-aware model learning which improves the performance of MBRL in challenging settings, such as small model capacity and the presence of distracting state dimensions. We analyze both MLE and value-aware approaches and demonstrate how they fail to account for exploration and the behavior of function approximation when learning value-aware models and highlight the additional goals that must be met to stabilize optimization in the deep learning setting. We verify our analysis by showing that our loss function is able to achieve high returns on the Mujoco benchmark suite while being more robust than maximum likelihood based approaches.

Amin Rakhsha, Andrew Wang, Mohammad Ghavamzadeh et al.

We introduce new planning and reinforcement learning algorithms for discounted MDPs that utilize an approximate model of the environment to accelerate the convergence of the value function. Inspired by the splitting approach in numerical linear algebra, we introduce Operator Splitting Value Iteration (OS-VI) for both Policy Evaluation and Control problems. OS-VI achieves a much faster convergence rate when the model is accurate enough. We also introduce a sample-based version of the algorithm called OS-Dyna. Unlike the traditional Dyna architecture, OS-Dyna still converges to the correct value function in presence of model approximation error.

Mete Kemertas, Allan D. Jepson, Amir-massoud Farahmand

We propose Mirror Descent Optimal Transport (MDOT), a novel method for solving discrete optimal transport (OT) problems with high precision, by unifying temperature annealing in entropic-regularized OT (EOT) with mirror descent techniques. In this framework, temperature annealing produces a sequence of EOT dual problems, whose solution gradually gets closer to the solution of the original OT problem. We solve each problem efficiently using a GPU-parallel nonlinear conjugate gradients algorithm (PNCG) that outperforms traditional Sinkhorn iterations under weak regularization. Moreover, our investigation also reveals that the theoretical convergence rate of Sinkhorn iterations can exceed existing non-asymptotic bounds when its stopping criterion is tuned in a manner analogous to MDOT. Our comprehensive ablation studies of MDOT-PNCG affirm its robustness across a wide range of algorithmic parameters. Benchmarking on 24 problem sets of size $n=4096$ in a GPU environment demonstrate that our method attains high-precision, feasible solutions significantly faster than a representative set of existing OT solvers, including accelerated gradient methods and advanced Sinkhorn variants, in both wall-clock time and number of operations. Empirical convergence rates range between $O(n^2 \varepsilon^{-1/4})$ and $O(n^2 \varepsilon^{-1})$, where $\varepsilon$ is the optimality gap. For problem sizes up to $n=16384$, the empirical runtime scales as $O(n^2)$ for moderate precision and as $O(n^{5/2})$ at worst for high precision. These findings establish MDOT-PNCG as a compelling alternative to current OT solvers, particularly in challenging weak-regularization regimes.

Amin Rakhsha, Mete Kemertas, Mohammad Ghavamzadeh et al.

We propose and theoretically analyze an approach for planning with an approximate model in reinforcement learning that can reduce the adverse impact of model error. If the model is accurate enough, it accelerates the convergence to the true value function too. One of its key components is the MaxEnt Model Correction (MoCo) procedure that corrects the model's next-state distributions based on a Maximum Entropy density estimation formulation. Based on MoCo, we introduce the Model Correcting Value Iteration (MoCoVI) algorithm, and its sampled-based variant MoCoDyna. We show that MoCoVI and MoCoDyna's convergence can be much faster than the conventional model-free algorithms. Unlike traditional model-based algorithms, MoCoVI and MoCoDyna effectively utilize an approximate model and still converge to the correct value function.

Tyler Kastner, Murat A. Erdogdu, Amir-massoud Farahmand

We consider the problem of learning models for risk-sensitive reinforcement learning. We theoretically demonstrate that proper value equivalence, a method of learning models which can be used to plan optimally in the risk-neutral setting, is not sufficient to plan optimally in the risk-sensitive setting. We leverage distributional reinforcement learning to introduce two new notions of model equivalence, one which is general and can be used to plan for any risk measure, but is intractable; and a practical variation which allows one to choose which risk measures they may plan optimally for. We demonstrate how our framework can be used to augment any model-free risk-sensitive algorithm, and provide both tabular and large-scale experiments to demonstrate its ability.

Avery Ma, Amir-massoud Farahmand, Yangchen Pan et al.

Neural networks are susceptible to adversarial perturbations that are transferable across different models. In this paper, we introduce a novel model alignment technique aimed at improving a given source model's ability in generating transferable adversarial perturbations. During the alignment process, the parameters of the source model are fine-tuned to minimize an alignment loss. This loss measures the divergence in the predictions between the source model and another, independently trained model, referred to as the witness model. To understand the effect of model alignment, we conduct a geometric analysis of the resulting changes in the loss landscape. Extensive experiments on the ImageNet dataset, using a variety of model architectures, demonstrate that perturbations generated from aligned source models exhibit significantly higher transferability than those from the original source model.

Avery Ma, Yangchen Pan, Amir-massoud Farahmand

Stochastic gradient descent (SGD) and adaptive gradient methods, such as Adam and RMSProp, have been widely used in training deep neural networks. We empirically show that while the difference between the standard generalization performance of models trained using these methods is small, those trained using SGD exhibit far greater robustness under input perturbations. Notably, our investigation demonstrates the presence of irrelevant frequencies in natural datasets, where alterations do not affect models' generalization performance. However, models trained with adaptive methods show sensitivity to these changes, suggesting that their use of irrelevant frequencies can lead to solutions sensitive to perturbations. To better understand this difference, we study the learning dynamics of gradient descent (GD) and sign gradient descent (signGD) on a synthetic dataset that mirrors natural signals. With a three-dimensional input space, the models optimized with GD and signGD have standard risks close to zero but vary in their adversarial risks. Our result shows that linear models' robustness to $\ell_2$-norm bounded changes is inversely proportional to the model parameters' weight norm: a smaller weight norm implies better robustness. In the context of deep learning, our experiments show that SGD-trained neural networks have smaller Lipschitz constants, explaining the better robustness to input perturbations than those trained with adaptive gradient methods.

Jongmin Lee, Amin Rakhsha, Ernest K. Ryu et al.

The Value Iteration (VI) algorithm is an iterative procedure to compute the value function of a Markov decision process, and is the basis of many reinforcement learning (RL) algorithms as well. As the error convergence rate of VI as a function of iteration $k$ is $O(γ^k)$, it is slow when the discount factor $γ$ is close to $1$. To accelerate the computation of the value function, we propose Deflated Dynamics Value Iteration (DDVI). DDVI uses matrix splitting and matrix deflation techniques to effectively remove (deflate) the top $s$ dominant eigen-structure of the transition matrix $\mathcal{P}^π$. We prove that this leads to a $\tilde{O}(γ^k |λ_{s+1}|^k)$ convergence rate, where $λ_{s+1}$is $(s+1)$-th largest eigenvalue of the dynamics matrix. We then extend DDVI to the RL setting and present Deflated Dynamics Temporal Difference (DDTD) algorithm. We empirically show the effectiveness of the proposed algorithms.

Mark Bedaywi, Amin Rakhsha, Amir-massoud Farahmand

Long-horizon tasks, which have a large discount factor, pose a challenge for most conventional reinforcement learning (RL) algorithms. Algorithms such as Value Iteration and Temporal Difference (TD) learning have a slow convergence rate and become inefficient in these tasks. When the transition distributions are given, PID VI was recently introduced to accelerate the convergence of Value Iteration using ideas from control theory. Inspired by this, we introduce PID TD Learning and PID Q-Learning algorithms for the RL setting, in which only samples from the environment are available. We give a theoretical analysis of the convergence of PID TD Learning and its acceleration compared to the conventional TD Learning. We also introduce a method for adapting PID gains in the presence of noise and empirically verify its effectiveness.

Claas A Voelcker, Arash Ahmadian, Romina Abachi et al.

The idea of decision-aware model learning, that models should be accurate where it matters for decision-making, has gained prominence in model-based reinforcement learning. While promising theoretical results have been established, the empirical performance of algorithms leveraging a decision-aware loss has been lacking, especially in continuous control problems. In this paper, we present a study on the necessary components for decision-aware reinforcement learning models and we showcase design choices that enable well-performing algorithms. To this end, we provide a theoretical and empirical investigation into algorithmic ideas in the field. We highlight that empirical design decisions established in the MuZero line of works, most importantly the use of a latent model, are vital to achieving good performance for related algorithms. Furthermore, we show that the MuZero loss function is biased in stochastic environments and establish that this bias has practical consequences. Building on these findings, we present an overview of which decision-aware loss functions are best used in what empirical scenarios, providing actionable insights to practitioners in the field.

Marcel Hussing, Claas Voelcker, Igor Gilitschenski et al.

We show that deep reinforcement learning algorithms can retain their ability to learn without resetting network parameters in settings where the number of gradient updates greatly exceeds the number of environment samples by combatting value function divergence. Under large update-to-data ratios, a recent study by Nikishin et al. (2022) suggested the emergence of a primacy bias, in which agents overfit early interactions and downplay later experience, impairing their ability to learn. In this work, we investigate the phenomena leading to the primacy bias. We inspect the early stages of training that were conjectured to cause the failure to learn and find that one fundamental challenge is a long-standing acquaintance: value function divergence. Overinflated Q-values are found not only on out-of-distribution but also in-distribution data and can be linked to overestimation on unseen action prediction propelled by optimizer momentum. We employ a simple unit-ball normalization that enables learning under large update ratios, show its efficacy on the widely used dm_control suite, and obtain strong performance on the challenging dog tasks, competitive with model-based approaches. Our results question, in parts, the prior explanation for sub-optimal learning due to overfitting early data.

Claas A Voelcker, Marcel Hussing, Eric Eaton et al.

Building deep reinforcement learning (RL) agents that find a good policy with few samples has proven notoriously challenging. To achieve sample efficiency, recent work has explored updating neural networks with large numbers of gradient steps for every new sample. While such high update-to-data (UTD) ratios have shown strong empirical performance, they also introduce instability to the training process. Previous approaches need to rely on periodic neural network parameter resets to address this instability, but restarting the training process is infeasible in many real-world applications and requires tuning the resetting interval. In this paper, we focus on one of the core difficulties of stable training with limited samples: the inability of learned value functions to generalize to unobserved on-policy actions. We mitigate this issue directly by augmenting the off-policy RL training process with a small amount of data generated from a learned world model. Our method, Model-Augmented Data for TD Learning (MAD-TD), uses small amounts of generated data to stabilize high UTD training and achieve competitive performance on the most challenging tasks in the DeepMind control suite. Our experiments further highlight the importance of employing a good model to generate data, MAD-TD's ability to combat value overestimation, and its practical stability gains for continued learning.

Avery Ma, Yangchen Pan, Amir-massoud Farahmand

Many-shot jailbreaking circumvents the safety alignment of LLMs by exploiting their ability to process long input sequences. To achieve this, the malicious target prompt is prefixed with hundreds of fabricated conversational exchanges between the user and the model. These exchanges are randomly sampled from a pool of unsafe question-answer pairs, making it appear as though the model has already complied with harmful instructions. In this paper, we present PANDAS: a hybrid technique that improves many-shot jailbreaking by modifying these fabricated dialogues with Positive Affirmations, Negative Demonstrations, and an optimized Adaptive Sampling method tailored to the target prompt's topic. We also introduce ManyHarm, a dataset of harmful question-answer pairs, and demonstrate through extensive experiments that PANDAS significantly outperforms baseline methods in long-context scenarios. Through attention analysis, we provide insights into how long-context vulnerabilities are exploited and show how PANDAS further improves upon many-shot jailbreaking.

Mete Kemertas, Amir-massoud Farahmand, Allan D. Jepson

Developing a contemporary optimal transport (OT) solver requires navigating trade-offs among several critical requirements: GPU parallelization, scalability to high-dimensional problems, theoretical convergence guarantees, empirical performance in terms of precision versus runtime, and numerical stability in practice. With these challenges in mind, we introduce a specialized truncated Newton algorithm for entropic-regularized OT. In addition to proving that locally quadratic convergence is possible without assuming a Lipschitz Hessian, we provide strategies to maximally exploit the high rate of local convergence in practice. Our GPU-parallel algorithm exhibits exceptionally favorable runtime performance, achieving high precision orders of magnitude faster than many existing alternatives. This is evidenced by wall-clock time experiments on 24 problem sets (12 datasets $\times$ 2 cost functions). The scalability of the algorithm is showcased on an extremely large OT problem with $n \approx 10^6$, solved approximately under weak entopric regularization.

Amin Rakhsha, Kanika Madan, Tianyu Zhang et al.

Sampling multiple outputs from a Large Language Model (LLM) and selecting the most frequent (Self-consistency) or highest-scoring (Best-of-N) candidate is a popular approach to achieve higher accuracy in tasks with discrete final answers. Best-of-N (BoN) selects the output with the highest reward, and with perfect rewards, it often achieves near-perfect accuracy. With imperfect rewards from reward models, however, BoN fails to reliably find the correct answer and its performance degrades drastically. We consider the distribution of BoN's outputs and highlight that, although the correct answer does not usually have a probability close to one under imperfect rewards, it is often the most likely outcome. This suggests that the mode of this distribution can be more reliably correct than a sample from it. Based on this idea, we propose Majority-of-the-Bests (MoB), a novel selection mechanism that estimates the output distribution of BoN via bootstrapping and selects its mode. Experimental results across five benchmarks, three different base LLMs, and two reward models demonstrate consistent improvements over BoN in 25 out of 30 setups. We also provide theoretical results for the consistency of the bootstrapping. MoB serves as a simple, yet strong alternative to BoN and self-consistency, and more broadly, motivates further research in more nuanced selection mechanisms.

Claas Voelcker, Axel Brunnbauer, Marcel Hussing et al.

Score-function based methods for policy learning, such as REINFORCE and PPO, have delivered strong results in game-playing and robotics, yet their high variance often undermines training stability. Using pathwise policy gradients, i.e. computing a derivative by differentiating the objective function, alleviates the variance issues. However, they require an accurate action-conditioned value function, which is notoriously hard to learn without relying on replay buffers for reusing past off-policy data. We present an on-policy algorithm that trains Q-value models purely from on-policy trajectories, unlocking the possibility of using pathwise policy updates in the context of on-policy learning. We show how to combine stochastic policies for exploration with constrained updates for stable training, and evaluate important architectural components that stabilize value function learning. The result, Relative Entropy Pathwise Policy Optimization (REPPO), is an efficient on-policy algorithm that combines the stability of pathwise policy gradients with the simplicity and minimal memory footprint of standard on-policy learning. Compared to state-of-the-art on two standard GPU-parallelized benchmarks, REPPO provides strong empirical performance at superior sample efficiency, wall-clock time, memory footprint, and hyperparameter robustness.

Claas Voelcker, Anastasiia Pedan, Arash Ahmadian et al.

The idea of value-aware model learning, that models should produce accurate value estimates, has gained prominence in model-based reinforcement learning. The MuZero loss, which penalizes a model's value function prediction compared to the ground-truth value function, has been utilized in several prominent empirical works in the literature. However, theoretical investigation into its strengths and weaknesses is limited. In this paper, we analyze the family of value-aware model learning losses, which includes the popular MuZero loss. We show that these losses, as normally used, are uncalibrated surrogate losses, which means that they do not always recover the correct model and value function. Building on this insight, we propose corrections to solve this issue. Furthermore, we investigate the interplay between the loss calibration, latent model architectures, and auxiliary losses that are commonly employed when training MuZero-style agents. We show that while deterministic models can be sufficient to predict accurate values, learning calibrated stochastic models is still advantageous.

Claas Voelcker, Tyler Kastner, Igor Gilitschenski et al.

We investigate the impact of auxiliary learning tasks such as observation reconstruction and latent self-prediction on the representation learning problem in reinforcement learning. We also study how they interact with distractions and observation functions in the MDP. We provide a theoretical analysis of the learning dynamics of observation reconstruction, latent self-prediction, and TD learning in the presence of distractions and observation functions under linear model assumptions. With this formalization, we are able to explain why latent-self prediction is a helpful \emph{auxiliary task}, while observation reconstruction can provide more useful features when used in isolation. Our empirical analysis shows that the insights obtained from our learning dynamics framework predicts the behavior of these loss functions beyond the linear model assumption in non-linear neural networks. This reinforces the usefulness of the linear model framework not only for theoretical analysis, but also practical benefit for applied problems.

Erfan Pirmorad, Faraz Khoshbakhtian, Farnam Mansouri et al.

In many areas, such as the physical sciences, life sciences, and finance, control approaches are used to achieve a desired goal in complex dynamical systems governed by differential equations. In this work we formulate the problem of controlling stochastic partial differential equations (SPDE) as a reinforcement learning problem. We present a learning-based, distributed control approach for online control of a system of SPDEs with high dimensional state-action space using deep deterministic policy gradient method. We tested the performance of our method on the problem of controlling the stochastic Burgers' equation, describing a turbulent fluid flow in an infinitely large domain.

Rodrigo Toro Icarte, Richard Valenzano, Toryn Q. Klassen et al.

Reinforcement Learning (RL) agents typically learn memoryless policies---policies that only consider the last observation when selecting actions. Learning memoryless policies is efficient and optimal in fully observable environments. However, some form of memory is necessary when RL agents are faced with partial observability. In this paper, we study a lightweight approach to tackle partial observability in RL. We provide the agent with an external memory and additional actions to control what, if anything, is written to the memory. At every step, the current memory state is part of the agent's observation, and the agent selects a tuple of actions: one action that modifies the environment and another that modifies the memory. When the external memory is sufficiently expressive, optimal memoryless policies yield globally optimal solutions. Unfortunately, previous attempts to use external memory in the form of binary memory have produced poor results in practice. Here, we investigate alternative forms of memory in support of learning effective memoryless policies. Our novel forms of memory outperform binary and LSTM-based memory in well-established partially observable domains.

Yangchen Pan, Jincheng Mei, Amir-massoud Farahmand et al.

Prioritized Experience Replay (ER) has been empirically shown to improve sample efficiency across many domains and attracted great attention; however, there is little theoretical understanding of why such prioritized sampling helps and its limitations. In this work, we take a deep look at the prioritized ER. In a supervised learning setting, we show the equivalence between the error-based prioritized sampling method for mean squared error and uniform sampling for cubic power loss. We then provide theoretical insight into why it improves convergence rate upon uniform sampling during early learning. Based on the insight, we further point out two limitations of the prioritized ER method: 1) outdated priorities and 2) insufficient coverage of the sample space. To mitigate the limitations, we propose our model-based stochastic gradient Langevin dynamics sampling method. We show that our method does provide states distributed close to an ideal prioritized sampling distribution estimated by the brute-force method, which does not suffer from the two limitations. We conduct experiments on both discrete and continuous control problems to show our approach's efficacy and examine the practical implication of our method in an autonomous driving application.

Avery Ma, Fartash Faghri, Nicolas Papernot et al.

Adversarial training is a common approach to improving the robustness of deep neural networks against adversarial examples. In this work, we propose a novel regularization approach as an alternative. To derive the regularizer, we formulate the adversarial robustness problem under the robust optimization framework and approximate the loss function using a second-order Taylor series expansion. Our proposed second-order adversarial regularizer (SOAR) is an upper bound based on the Taylor approximation of the inner-max in the robust optimization objective. We empirically show that the proposed method significantly improves the robustness of networks against the $\ell_\infty$ and $\ell_2$ bounded perturbations generated using cross-entropy-based PGD on CIFAR-10 and SVHN.

Romina Abachi, Mohammad Ghavamzadeh, Amir-massoud Farahmand

This paper considers the problem of learning a model in model-based reinforcement learning (MBRL). We examine how the planning module of an MBRL algorithm uses the model, and propose that the model learning module should incorporate the way the planner is going to use the model. This is in contrast to conventional model learning approaches, such as those based on maximum likelihood estimate, that learn a predictive model of the environment without explicitly considering the interaction of the model and the planner. We focus on policy gradient type of planning algorithms and derive new loss functions for model learning that incorporate how the planner uses the model. We call this approach Policy-Aware Model Learning (PAML). We theoretically analyze a generic model-based policy gradient algorithm and provide a convergence guarantee for the optimized policy. We also empirically evaluate PAML on some benchmark problems, showing promising results.

Yangchen Pan, Ehsan Imani, Martha White et al.

For multi-valued functions---such as when the conditional distribution on targets given the inputs is multi-modal---standard regression approaches are not always desirable because they provide the conditional mean. Modal regression algorithms address this issue by instead finding the conditional mode(s). Most, however, are nonparametric approaches and so can be difficult to scale. Further, parametric approximators, like neural networks, facilitate learning complex relationships between inputs and targets. In this work, we propose a parametric modal regression algorithm. We use the implicit function theorem to develop an objective, for learning a joint function over inputs and targets. We empirically demonstrate on several synthetic problems that our method (i) can learn multi-valued functions and produce the conditional modes, (ii) scales well to high-dimensional inputs, and (iii) can even be more effective for certain uni-modal problems, particularly for high-frequency functions. We demonstrate that our method is competitive in a real-world modal regression problem and two regular regression datasets.

Yangchen Pan, Jincheng Mei, Amir-massoud Farahmand

Model-based reinforcement learning has been empirically demonstrated as a successful strategy to improve sample efficiency. In particular, Dyna is an elegant model-based architecture integrating learning and planning that provides huge flexibility of using a model. One of the most important components in Dyna is called search-control, which refers to the process of generating state or state-action pairs from which we query the model to acquire simulated experiences. Search-control is critical in improving learning efficiency. In this work, we propose a simple and novel search-control strategy by searching high frequency regions of the value function. Our main intuition is built on Shannon sampling theorem from signal processing, which indicates that a high frequency signal requires more samples to reconstruct. We empirically show that a high frequency function is more difficult to approximate. This suggests a search-control strategy: we should use states from high frequency regions of the value function to query the model to acquire more samples. We develop a simple strategy to locally measure the frequency of a function by gradient and hessian norms, and provide theoretical justification for this approach. We then apply our strategy to search-control in Dyna, and conduct experiments to show its property and effectiveness on benchmark domains.

Yangchen Pan, Hengshuai Yao, Amir-massoud Farahmand et al.

Dyna is an architecture for model-based reinforcement learning (RL), where simulated experience from a model is used to update policies or value functions. A key component of Dyna is search-control, the mechanism to generate the state and action from which the agent queries the model, which remains largely unexplored. In this work, we propose to generate such states by using the trajectory obtained from Hill Climbing (HC) the current estimate of the value function. This has the effect of propagating value from high-value regions and of preemptively updating value estimates of the regions that the agent is likely to visit next. We derive a noisy projected natural gradient algorithm for hill climbing, and highlight a connection to Langevin dynamics. We provide an empirical demonstration on four classical domains that our algorithm, HC-Dyna, can obtain significant sample efficiency improvements. We study the properties of different sampling distributions for search-control, and find that there appears to be a benefit specifically from using the samples generated by climbing on current value estimates from low-value to high-value region.

Mohamed Akrout, Amir-massoud Farahmand, Tory Jarmain et al.

We present a visual symptom checker that combines a pre-trained Convolutional Neural Network (CNN) with a Reinforcement Learning (RL) agent as a Question Answering (QA) model. This method increases the classification confidence and accuracy of the visual symptom checker, and decreases the average number of questions asked to narrow down the differential diagnosis. A Deep Q-Network (DQN)-based RL agent learns how to ask the patient about the presence of symptoms in order to maximize the probability of correctly identifying the underlying condition. The RL agent uses the visual information provided by CNN in addition to the answers to the asked questions to guide the QA system. We demonstrate that the RL-based approach increases the accuracy more than 20% compared to the CNN-only approach, which only uses the visual information to predict the condition. Moreover, the increased accuracy is up to 10% compared to the approach that uses the visual information provided by CNN along with a conventional decision tree-based QA system. We finally show that the RL-based approach not only outperforms the decision tree-based approach, but also narrows down the diagnosis faster in terms of the average number of asked questions.

Mohamed Akrout, Amir-massoud Farahmand, Tory Jarmain

We present a skin condition classification methodology based on a sequential pipeline of a pre-trained Convolutional Neural Network (CNN) and a Question Answering (QA) model. This method enables us to not only increase the classification confidence and accuracy of the deployed CNN system, but also enables the emulation of the conventional approach of doctors asking the relevant questions in refining the ultimate diagnosis and differential. By combining the CNN output in the form of classification probabilities as a prior to the QA model and the image textual description, we greedily ask the best symptom that maximizes the information gain over symptoms. We demonstrate that combining the QA model with the CNN increases the accuracy up to 10% as compared to the CNN alone, and more than 30% as compared to the QA model alone.

Yangchen Pan, Amir-massoud Farahmand, Martha White et al.

Recent work has shown that reinforcement learning (RL) is a promising approach to control dynamical systems described by partial differential equations (PDE). This paper shows how to use RL to tackle more general PDE control problems that have continuous high-dimensional action spaces with spatial relationship among action dimensions. In particular, we propose the concept of action descriptors, which encode regularities among spatially-extended action dimensions and enable the agent to control high-dimensional action PDEs. We provide theoretical evidence suggesting that this approach can be more sample efficient compared to a conventional approach that treats each action dimension separately and does not explicitly exploit the spatial regularity of the action space. The action descriptor approach is then used within the deep deterministic policy gradient algorithm. Experiments on two PDE control problems, with up to 256-dimensional continuous actions, show the advantage of the proposed approach over the conventional one.

Kota Hara, Ming-Yu Liu, Oncel Tuzel et al.

We propose augmenting deep neural networks with an attention mechanism for the visual object detection task. As perceiving a scene, humans have the capability of multiple fixation points, each attended to scene content at different locations and scales. However, such a mechanism is missing in the current state-of-the-art visual object detection methods. Inspired by the human vision system, we propose a novel deep network architecture that imitates this attention mechanism. As detecting objects in an image, the network adaptively places a sequence of glimpses of different shapes at different locations in the image. Evidences of the presence of an object and its location are extracted from these glimpses, which are then fused for estimating the object class and bounding box coordinates. Due to lacks of ground truth annotations of the visual attention mechanism, we train our network using a reinforcement learning algorithm with policy gradients. Experiment results on standard object detection benchmarks show that the proposed network consistently outperforms the baseline networks that does not model the attention mechanism.

Mouhacine Benosman, Amir-massoud Farahmand, Meng Xia

We study in this paper the problem of adaptive trajectory tracking control for a class of nonlinear systems with parametric uncertainties. We propose to use a modular approach, where we first design a robust nonlinear state feedback which renders the closed loop input-to-state stable (ISS), where the input is considered to be the estimation error of the uncertain parameters, and the state is considered to be the closed-loop output tracking error. Next, we augment this robust ISS controller with a model-free learning algorithm to estimate the model uncertainties. We implement this method with two different learning approaches. The first one is a model-free multi-parametric extremum seeking (MES) method and the second is a Bayesian optimization-based method called Gaussian Process Upper Confidence Bound (GP-UCB). The combination of the ISS feedback and the learning algorithms gives a learning-based modular indirect adaptive controller. We show the efficiency of this approach on a two-link robot manipulator example.

Amir-massoud Farahmand, Doina Precup, André M. S. Barreto et al.

Tackling large approximate dynamic programming or reinforcement learning problems requires methods that can exploit regularities, or intrinsic structure, of the problem in hand. Most current methods are geared towards exploiting the regularities of either the value function or the policy. We introduce a general classification-based approximate policy iteration (CAPI) framework, which encompasses a large class of algorithms that can exploit regularities of both the value function and the policy space, depending on what is advantageous. This framework has two main components: a generic value function estimator and a classifier that learns a policy based on the estimated value function. We establish theoretical guarantees for the sample complexity of CAPI-style algorithms, which allow the policy evaluation step to be performed by a wide variety of algorithms (including temporal-difference-style methods), and can handle nonparametric representations of policies. Our bounds on the estimation error of the performance loss are tighter than existing results. We also illustrate this approach empirically on several problems, including a large HIV control task.

Mahdi Milani Fard, Yuri Grinberg, Amir-massoud Farahmand et al.

We address the problem of automatic generation of features for value function approximation. Bellman Error Basis Functions (BEBFs) have been shown to improve the error of policy evaluation with function approximation, with a convergence rate similar to that of value iteration. We propose a simple, fast and robust algorithm based on random projections to generate BEBFs for sparse feature spaces. We provide a finite sample analysis of the proposed method, and prove that projections logarithmic in the dimension of the original space are enough to guarantee contraction in the error. Empirical results demonstrate the strength of this method.