[論文レビュー] SoftAdapt: Techniques for Adaptive Loss Weighting of Neural Networks with Multi-Part Loss Functions

A. Ali Heydari, Craig A. Thompson+1 more|arXiv (Cornell University)|Dec 27, 2019

Advanced Neural Network Applications参考文献 32被引用数 65

ひとこと要約

SoftAdaptは、各損失部分の最近の変化率に依存するソフトマックス風の方式を用いて、マルチパート損失の適応的重み付けを導入し、 manual tuning なしで収束を改善します。

ABSTRACT

Adaptive loss function formulation is an active area of research and has gained a great deal of popularity in recent years, following the success of deep learning. However, existing frameworks of adaptive loss functions often suffer from slow convergence and poor choice of weights for the loss components. Traditionally, the elements of a multi-part loss function are weighted equally or their weights are determined through heuristic approaches that yield near-optimal (or sub-optimal) results. To address this problem, we propose a family of methods, called SoftAdapt, that dynamically change function weights for multi-part loss functions based on live performance statistics of the component losses. SoftAdapt is mathematically intuitive, computationally efficient and straightforward to implement. In this paper, we present the mathematical formulation and pseudocode for SoftAdapt, along with results from applying our methods to image reconstruction (Sparse Autoencoders) and synthetic data generation (Introspective Variational Autoencoders).

研究の動機と目的

• ニューラルネットワークにおける複数の損失成分のバランス取りの課題に動機づけ、対処する。
• トレーニング中に損失項の重みを適応させる、一般的で高速かつ最適化手法と互換性のある方法を提案する。
• 適応的重み付けが、タスクを超えて固定重みやヒューリスティックに選択された重みより優れることを示す。
• Autoencoder、VAE、および勾配降下最適化のベンチマークへの適用性を示す。

提案手法

• multi-part lossを F(x)=sum_k f_k(x)として定式化し、重み付き勾配方向 h^i = sum_k alpha_k^i grad f_k(x^i) を定義する。
• 各 f_k の短期変化率として成分別性能率 s_k^i を算出する。
• SoftAdapt のVariant（Original）を用いて s^i の softmax に基づき重み alpha^i を計算する。
• Loss Weighted variantを用いて、alpha_k^i を現在の損失 f_k^iでスケーリングする。
• オプションとして、成分間の識別を鋭くするために rate ベクトルを正規化する。
• 任意の勾配降下型最適化器と統合可能な SoftAdapt およびその variants の疑似コードを提供する。

実験結果

リサーチクエスチョン

• RQ1 Adaptive weighting of loss components can improve training efficiency and outcomes over fixed equal weights?
• RQ2 How do different SoftAdapt variants (Original, Loss Weighted, Normalized) affect convergence across tasks and loss scales?
• RQ3 Is SoftAdapt compatible with common optimizers and architectures without substantial overhead?
• RQ4 How does adaptive weighting impact performance in autoencoders and VAEs compared to fixed heuristics?

主な発見

• SoftAdapt can yield faster convergence than fixed weights on benchmark optimization problems like Rosenbrock and Beale’s functions.
• In IntroVAE experiments, SoftAdapt-adaptive weights improve SSIM and PSNR metrics versus fixed weights while preserving similar training time.
• In Sparse Autoencoder experiments, SoftAdapt dynamically adjusting lambda improves reconstruction quality and classification performance compared to a fixed optimal lambda found by grid search.
• Across tasks, the adaptive weighting approach reduces the need for prior hyperparameter tuning and grid searches.
• The method is compatible with Adam and other gradient-based optimizers and is simple to implement as an add-on.