Get our free extension to see links to code for papers anywhere online!Free add-on: code for papers everywhere!Free add-on: See code for papers anywhere!

We focus on the task of learning the value function in the reinforcement learning (RL) setting. This task is often solved by updating a pair of online and target networks while ensuring that the parameters of these two networks are equivalent. We propose Lookahead-Replicate (LR), a new value-function approximation algorithm that is agnostic to this parameter-space equivalence. Instead, the LR algorithm is designed to maintain an equivalence between the two networks in the function space. This value-based equivalence is obtained by employing a new target-network update. We show that LR leads to a convergent behavior in learning the value function. We also present empirical results demonstrating that LR-based target-network updates significantly improve deep RL on the Atari benchmark.

Via

Kaan Ozkara, Can Karakus, Parameswaran Raman, Mingyi Hong, Shoham Sabach, Branislav Kveton, Volkan Cevher

Since Adam was introduced, several novel adaptive optimizers for deep learning have been proposed. These optimizers typically excel in some tasks but may not outperform Adam uniformly across all tasks. In this work, we introduce Meta-Adaptive Optimizers (MADA), a unified optimizer framework that can generalize several known optimizers and dynamically learn the most suitable one during training. The key idea in MADA is to parameterize the space of optimizers and search through it using hyper-gradient descent. Numerical results suggest that MADA is robust against sub-optimally tuned hyper-parameters, and outperforms Adam, Lion, and Adan with their default hyper-parameters, often even with optimized hyper-parameters. We also propose AVGrad, a variant of AMSGrad where the maximum operator is replaced with averaging, and observe that it performs better within MADA. Finally, we provide a convergence analysis to show that interpolation of optimizers (specifically, AVGrad and Adam) can improve their error bounds (up to constants), hinting at an advantage for meta-optimizers.

Via

Second-order optimization methods, such as cubic regularized Newton methods, are known for their rapid convergence rates; nevertheless, they become impractical in high-dimensional problems due to their substantial memory requirements and computational costs. One promising approach is to execute second-order updates within a lower-dimensional subspace, giving rise to subspace second-order methods. However, the majority of existing subspace second-order methods randomly select subspaces, consequently resulting in slower convergence rates depending on the problem's dimension $d$. In this paper, we introduce a novel subspace cubic regularized Newton method that achieves a dimension-independent global convergence rate of ${O}\left(\frac{1}{mk}+\frac{1}{k^2}\right)$ for solving convex optimization problems. Here, $m$ represents the subspace dimension, which can be significantly smaller than $d$. Instead of adopting a random subspace, our primary innovation involves performing the cubic regularized Newton update within the Krylov subspace associated with the Hessian and the gradient of the objective function. This result marks the first instance of a dimension-independent convergence rate for a subspace second-order method. Furthermore, when specific spectral conditions of the Hessian are met, our method recovers the convergence rate of a full-dimensional cubic regularized Newton method. Numerical experiments show our method converges faster than existing random subspace methods, especially for high-dimensional problems.

Via

The full potential of large pretrained models remains largely untapped in control domains like robotics. This is mainly because of the scarcity of data and the computational challenges associated with training or fine-tuning these large models for such applications. Prior work mainly emphasizes effective pretraining of large models for decision-making, with little exploration into how to perform data-efficient continual adaptation of these models for new tasks. Recognizing these constraints, we introduce TAIL (Task-specific Adapters for Imitation Learning), a framework for efficient adaptation to new control tasks. Inspired by recent advancements in parameter-efficient fine-tuning in language domains, we explore efficient fine-tuning techniques -- e.g., Bottleneck Adapters, P-Tuning, and Low-Rank Adaptation (LoRA) -- in TAIL to adapt large pretrained models for new tasks with limited demonstration data. Our extensive experiments in large-scale language-conditioned manipulation tasks comparing prevalent parameter-efficient fine-tuning techniques and adaptation baselines suggest that TAIL with LoRA can achieve the best post-adaptation performance with only 1\% of the trainable parameters of full fine-tuning, while avoiding catastrophic forgetting and preserving adaptation plasticity in continual learning settings.

Via

In this paper, we propose the Bi-Sub-Gradient (Bi-SG) method, which is a generalization of the classical sub-gradient method to the setting of convex bi-level optimization problems. This is a first-order method that is very easy to implement in the sense that it requires only a computation of the associated proximal mapping or a sub-gradient of the outer non-smooth objective function, in addition to a proximal gradient step on the inner optimization problem. We show, under very mild assumptions, that Bi-SG tackles bi-level optimization problems and achieves sub-linear rates both in terms of the inner and outer objective functions. Moreover, if the outer objective function is additionally strongly convex (still could be non-smooth), the outer rate can be improved to a linear rate. Last, we prove that the distance of the generated sequence to the set of optimal solutions of the bi-level problem converges to zero.

Via

We study the convergence behavior of the celebrated temporal-difference (TD) learning algorithm. By looking at the algorithm through the lens of optimization, we first argue that TD can be viewed as an iterative optimization algorithm where the function to be minimized changes per iteration. By carefully investigating the divergence displayed by TD on a classical counter example, we identify two forces that determine the convergent or divergent behavior of the algorithm. We next formalize our discovery in the linear TD setting with quadratic loss and prove that convergence of TD hinges on the interplay between these two forces. We extend this optimization perspective to prove convergence of TD in a much broader setting than just linear approximation and squared loss. Our results provide a theoretical explanation for the successful application of TD in reinforcement learning.

Via

We focus on the task of approximating the optimal value function in deep reinforcement learning. This iterative process is comprised of approximately solving a sequence of optimization problems where the objective function can change per iteration. The common approach to solving the problem is to employ modern variants of the stochastic gradient descent algorithm such as Adam. These optimizers maintain their own internal parameters such as estimates of the first and the second moment of the gradient, and update these parameters over time. Therefore, information obtained in previous iterations is being used to solve the optimization problem in the current iteration. We hypothesize that this can contaminate the internal parameters of the employed optimizer in situations where the optimization landscape of the previous iterations is quite different from the current iteration. To hedge against this effect, a simple idea is to reset the internal parameters of the optimizer when starting a new iteration. We empirically investigate this resetting strategy by employing various optimizers in conjunction with the Rainbow algorithm. We demonstrate that this simple modification unleashes the true potential of modern optimizers, and significantly improves the performance of deep RL on the Atari benchmark.

Via

Backtracking line-search is an old yet powerful strategy for finding better step size to be used in proximal gradient algorithms. The main principle is to locally find a simple convex upper bound of the objective function, which in turn controls the step size that is used. In case of inertial proximal gradient algorithms, the situation becomes much more difficult and usually leads to very restrictive rules on the extrapolation parameter. In this paper, we show that the extrapolation parameter can be controlled by locally finding also a simple concave lower bound of the objective function. This gives rise to a double convex-concave backtracking procedure which allows for an adaptive and optimal choice of both the step size and extrapolation parameters. We apply this procedure to the class of inertial Bregman proximal gradient methods, and prove that any sequence generated converges globally to critical points of the function at hand. Numerical experiments on a number of challenging non-convex problems in image processing and machine learning were conducted and show the power of combining inertial step and double backtracking strategy in achieving improved performances.

Via

Motivated by matrix recovery problems such as Robust Principal Component Analysis, we consider a general optimization problem of minimizing a smooth and strongly convex loss applied to the sum of two blocks of variables, where each block of variables is constrained or regularized individually. We present a novel Generalized Conditional Gradient method which is able to leverage the special structure of the problem to obtain faster convergence rates than those attainable via standard methods, under a variety of interesting assumptions. In particular, our method is appealing for matrix problems in which one of the blocks corresponds to a low-rank matrix and avoiding prohibitive full-rank singular value decompositions, which are required by most standard methods, is most desirable. Importantly, while our initial motivation comes from problems which originated in statistics, our analysis does not impose any statistical assumptions on the data.

Via