Nowadays, embedded systems are provided with cache memories that are large
enough to influence in both performance and energy consumption as never
occurred before in this kind of systems. In addition, the cache memory system
has been identified as a component that improves those metrics by adapting its
configuration according to the memory access patterns of the applications being
run. However, given that cache memories have many parameters which may be set
to a high number of different values, designers face to a wide and
time-consuming exploration space. In this paper we propose an optimization
framework based on Grammatical Evolution (GE) which is able to efficiently find
the best cache configurations for a given set of benchmark applications. This
metaheuristic allows an important reduction of the optimization runtime
obtaining good results in a low number of generations. Besides, this reduction
is also increased due to the efficient storage of evaluated caches. Moreover,
we selected GE because the plasticity of the grammar eases the creation of
phenotypes that form the call to the cache simulator required for the
evaluation of the different configurations. Experimental results for the
Mediabench suite show that our proposal is able to find cache configurations
that obtain an average improvement of $62\%$ versus a real world baseline
configuration.

本文提出了一种基于语法进化的优化框架，以高效地找到适用于给定基准应用程序的最佳缓存配置，对 Mediabench 套件进行实验，结果显示我们的提议能够找到适当的缓存配置，平均改善了 62％。

通过语法演化优化嵌入式系统的 L1 缓存

Optimizing L1 cache for embedded systems through grammatical evolution

Training large-scale image recognition models is computationally expensive.
This raises the question of whether there might be simple ways to improve the
test performance of an already trained model without having to re-train or
fine-tune it with new data. Here, we show that, surprisingly, this is indeed
possible. The key observation we make is that the layers of a deep network
close to the output layer contain independent, easily extractable
class-relevant information that is not contained in the output layer itself. We
propose to extract this extra class-relevant information using a simple
key-value cache memory to improve the classification performance of the model
at test time. Our cache memory is directly inspired by a similar cache model
previously proposed for language modeling (Grave et al., 2017). This cache
component does not require any training or fine-tuning; it can be applied to
any pre-trained model and, by properly setting only two hyper-parameters, leads
to significant improvements in its classification performance. Improvements are
observed across several architectures and datasets. In the cache component,
using features extracted from layers close to the output (but not from the
output layer itself) as keys leads to the largest improvements. Concatenating
features from multiple layers to form keys can further improve performance over
using single-layer features as keys. The cache component also has a
regularizing effect, a simple consequence of which is that it substantially
increases the robustness of models against adversarial attacks.

使用简单的键值缓存记忆提出了一种不需要重新训练或微调的方法来提高预训练模型的分类性能，并可以增强对抗攻击的鲁棒性。通过将靠近输出的层提取的特征作为键来形成键值，可以在大型图像识别模型中获得显著的性能提升。