arXiv subjects

Hongzi Mao

Publications and source records attributed to Hongzi Mao.

11 recordsLinked to original sources

arXiv2021

Flow-Loss: Learning Cardinality Estimates That Matter

Previous approaches to learned cardinality estimation have focused on improving average estimation error, but not all estimates matter equally. Since learned models inevitably make mistakes, the goal should be to improve the estimates that make the biggest difference to an optimizer. We introduce a new loss function, Flow-Loss, that explicitly optimizes for better query plans by approximating the optimizer's cost model and dynamic programming search algorithm with analytical functions. At the heart of Flow-Loss is a reduction of query optimization to a flow routing problem on a certain plan graph in which paths correspond to different query plans. To evaluate our approach, we introduce the Cardinality Estimation Benchmark, which contains the ground truth cardinalities for sub-plans of over 16K queries from 21 templates with up to 15 joins. We show that across different architectures and databases, a model trained with Flow-Loss improves the cost of plans (using the PostgreSQL cost model) and query runtimes despite having worse estimation accuracy than a model trained with Q-Error. When the test set queries closely match the training queries, both models improve performance significantly over PostgreSQL and are close to the optimal performance (using true cardinalities). However, the Q-Error trained model degrades significantly when evaluated on queries that are slightly different (e.g., similar but not identical query templates), while the Flow-Loss trained model generalizes better to such situations. For example, the Flow-Loss model achieves up to 1.5x better runtimes on unseen templates compared to the Q-Error model, despite leveraging the same model architecture and training data.

cs.DB↗

arXiv2020

Neural Rate Control for Video Encoding using Imitation Learning

In modern video encoders, rate control is a critical component and has been heavily engineered. It decides how many bits to spend to encode each frame, in order to optimize the rate-distortion trade-off over all video frames. This is a challenging constrained planning problem because of the complex dependency among decisions for different video frames and the bitrate constraint defined at the end of the episode. We formulate the rate control problem as a Partially Observable Markov Decision Process (POMDP), and apply imitation learning to learn a neural rate control policy. We demonstrate that by learning from optimal video encoding trajectories obtained through evolution strategies, our learned policy achieves better encoding efficiency and has minimal constraint violation. In addition to imitating the optimal actions, we find that additional auxiliary losses, data augmentation/refinement and inference-time policy improvements are critical for learning a good rate control policy. We evaluate the learned policy against the rate control policy in libvpx, a widely adopted open source VP9 codec library, in the two-pass variable bitrate (VBR) mode. We show that over a diverse set of real-world videos, our learned policy achieves 8.5% median bitrate reduction without sacrificing video quality.

cs.LG↗

arXiv2020

Real-world Video Adaptation with Reinforcement Learning

Client-side video players employ adaptive bitrate (ABR) algorithms to optimize user quality of experience (QoE). We evaluate recently proposed RL-based ABR methods in Facebook's web-based video streaming platform. Real-world ABR contains several challenges that requires customized designs beyond off-the-shelf RL algorithms -- we implement a scalable neural network architecture that supports videos with arbitrary bitrate encodings; we design a training method to cope with the variance resulting from the stochasticity in network conditions; and we leverage constrained Bayesian optimization for reward shaping in order to optimize the conflicting QoE objectives. In a week-long worldwide deployment with more than 30 million video streaming sessions, our RL approach outperforms the existing human-engineered ABR algorithms.

cs.NI↗

arXiv2020

Interpreting Deep Learning-Based Networking Systems

While many deep learning (DL)-based networking systems have demonstrated superior performance, the underlying Deep Neural Networks (DNNs) remain blackboxes and stay uninterpretable for network operators. The lack of interpretability makes DL-based networking systems prohibitive to deploy in practice. In this paper, we propose Metis, a framework that provides interpretability for two general categories of networking problems spanning local and global control. Accordingly, Metis introduces two different interpretation methods based on decision tree and hypergraph, where it converts DNN policies to interpretable rule-based controllers and highlight critical components based on analysis over hypergraph. We evaluate Metis over several state-of-the-art DL-based networking systems and show that Metis provides human-readable interpretations while preserving nearly no degradation in performance. We further present four concrete use cases of Metis, showcasing how Metis helps network operators to design, debug, deploy, and ad-hoc adjust DL-based networking systems.

cs.NI↗

arXiv2020

Learned Garbage Collection

Several programming languages use garbage collectors (GCs) to automatically manage memory for the programmer. Such collectors must decide when to look for unreachable objects to free, which can have a large performance impact on some applications. In this preliminary work, we propose a design for a learned garbage collector that autonomously learns over time when to perform collections. By using reinforcement learning, our design can incorporate user-defined reward functions, allowing an autonomous garbage collector to learn to optimize the exact metric the user desires (e.g., request latency or queries per second). We conduct an initial experimental study on a prototype, demonstrating that an approach based on tabular Q learning may be promising.

cs.PL↗

arXiv2020

Bao: Learning to Steer Query Optimizers

Query optimization remains one of the most challenging problems in data management systems. Recent efforts to apply machine learning techniques to query optimization challenges have been promising, but have shown few practical gains due to substantive training overhead, inability to adapt to changes, and poor tail performance. Motivated by these difficulties and drawing upon a long history of research in multi-armed bandits, we introduce Bao (the BAndit Optimizer). Bao takes advantage of the wisdom built into existing query optimizers by providing per-query optimization hints. Bao combines modern tree convolutional neural networks with Thompson sampling, a decades-old and well-studied reinforcement learning algorithm. As a result, Bao automatically learns from its mistakes and adapts to changes in query workloads, data, and schema. Experimentally, we demonstrate that Bao can quickly (an order of magnitude faster than previous approaches) learn strategies that improve end-to-end query execution performance, including tail latency. In cloud environments, we show that Bao can offer both reduced costs and better performance compared with a sophisticated commercial system.

cs.DB↗

arXiv2019

Learning Scheduling Algorithms for Data Processing Clusters

Efficiently scheduling data processing jobs on distributed compute clusters requires complex algorithms. Current systems, however, use simple generalized heuristics and ignore workload characteristics, since developing and tuning a scheduling policy for each workload is infeasible. In this paper, we show that modern machine learning techniques can generate highly-efficient policies automatically. Decima uses reinforcement learning (RL) and neural networks to learn workload-specific scheduling algorithms without any human instruction beyond a high-level objective such as minimizing average job completion time. Off-the-shelf RL techniques, however, cannot handle the complexity and scale of the scheduling problem. To build Decima, we had to develop new representations for jobs' dependency graphs, design scalable RL models, and invent RL training methods for dealing with continuous stochastic job arrivals. Our prototype integration with Spark on a 25-node cluster shows that Decima improves the average job completion time over hand-tuned scheduling heuristics by at least 21%, achieving up to 2x improvement during periods of high cluster load.

cs.LG↗

arXiv2019

Placeto: Learning Generalizable Device Placement Algorithms for Distributed Machine Learning

We present Placeto, a reinforcement learning (RL) approach to efficiently find device placements for distributed neural network training. Unlike prior approaches that only find a device placement for a specific computation graph, Placeto can learn generalizable device placement policies that can be applied to any graph. We propose two key ideas in our approach: (1) we represent the policy as performing iterative placement improvements, rather than outputting a placement in one shot; (2) we use graph embeddings to capture relevant information about the structure of the computation graph, without relying on node labels for indexing. These ideas allow Placeto to train efficiently and generalize to unseen graphs. Our experiments show that Placeto requires up to 6.1x fewer training steps to find placements that are on par with or better than the best placements found by prior approaches. Moreover, Placeto is able to learn a generalizable placement policy for any given family of graphs, which can then be used without any retraining to predict optimized placements for unseen graphs from the same family. This eliminates the large overhead incurred by prior RL approaches whose lack of generalizability necessitates re-training from scratch every time a new graph is to be placed.

cs.LG↗

arXiv2019

Neo: A Learned Query Optimizer

Query optimization is one of the most challenging problems in database systems. Despite the progress made over the past decades, query optimizers remain extremely complex components that require a great deal of hand-tuning for specific workloads and datasets. Motivated by this shortcoming and inspired by recent advances in applying machine learning to data management challenges, we introduce Neo (Neural Optimizer), a novel learning-based query optimizer that relies on deep neural networks to generate query executions plans. Neo bootstraps its query optimization model from existing optimizers and continues to learn from incoming queries, building upon its successes and learning from its failures. Furthermore, Neo naturally adapts to underlying data patterns and is robust to estimation errors. Experimental results demonstrate that Neo, even when bootstrapped from a simple optimizer like PostgreSQL, can learn a model that offers similar performance to state-of-the-art commercial optimizers, and in some cases even surpass them.

cs.DB↗

arXiv2019

Variance Reduction for Reinforcement Learning in Input-Driven Environments

We consider reinforcement learning in input-driven environments, where an exogenous, stochastic input process affects the dynamics of the system. Input processes arise in many applications, including queuing systems, robotics control with disturbances, and object tracking. Since the state dynamics and rewards depend on the input process, the state alone provides limited information for the expected future returns. Therefore, policy gradient methods with standard state-dependent baselines suffer high variance during training. We derive a bias-free, input-dependent baseline to reduce this variance, and analytically show its benefits over state-dependent baselines. We then propose a meta-learning approach to overcome the complexity of learning a baseline that depends on a long sequence of inputs. Our experimental results show that across environments from queuing systems, computer networks, and MuJoCo robotic locomotion, input-dependent baselines consistently improve training stability and result in better eventual policies.

cs.LG↗

arXiv2015

Fluctuation-response Relation Unifies Dynamical Behaviors in Neural Fields

Anticipation is a strategy used by neural fields to compensate for transmission and processing delays during the tracking of dynamical information, and can be achieved by slow, localized, inhibitory feedback mechanisms such as short-term synaptic depression, spike-frequency adaptation, or inhibitory feedback from other layers. Based on the translational symmetry of the mobile network states, we derive generic fluctuation-response relations, providing unified predictions that link their tracking behaviors in the presence of external stimuli to the intrinsic dynamics of the neural fields in their absence.

cond-mat.dis-nn↗