Deep neural networks (DNNs) excel in tasks like image recognition and natural language processing, but their increasing complexity complicates deployment in resource-constrained environments and increases susceptibility to adversarial attacks. While traditional pruning methods reduce model size, they often compromise the network's ability to withstand subtle perturbations. This paper challenges the conventional emphasis on weight importance scoring as the primary determinant of a pruned network's performance. Through extensive analysis, including experiments conducted on CIFAR, Tiny-ImageNet, and various network architectures, we demonstrate that effective fine-tuning plays a dominant role in enhancing both performance and adversarial robustness, often surpassing the impact of the chosen pruning criteria. To address this issue, we introduce Module Robust Sensitivity, a novel metric that adaptively adjusts the pruning ratio for each network layer based on its sensitivity to adversarial perturbations. By integrating this metric into the pruning process, we develop a stable algorithm that maintains accuracy and robustness simultaneously. Experimental results show that our approach enables the practical deployment of more robust and efficient neural networks.

It has become mainstream in computer vision and other machine learning domains to reuse backbone networks pre-trained on large datasets as preprocessors. Typically, the last layer is replaced by a shallow learning machine of sorts; the newly-added classification head and (optionally) deeper layers are fine-tuned on a new task. Due to its strong performance and simplicity, a common pre-trained backbone network is ResNet152.However, ResNet152 is relatively large and induces inference latency. In many cases, a compact and efficient backbone with similar performance would be preferable over a larger, slower one. This paper investigates techniques to reuse a pre-trained backbone with the objective of creating a smaller and faster model. Starting from a large ResNet152 backbone pre-trained on ImageNet, we first reduce it from 51 blocks to 5 blocks, reducing its number of parameters and FLOPs by more than 6 times, without significant performance degradation. Then, we split the model after 3 blocks into several branches, while preserving the same number of parameters and FLOPs, to create an ensemble of sub-networks to improve performance. Our experiments on a large benchmark of $40$ image classification datasets from various domains suggest that our techniques match the performance (if not better) of ``classical backbone fine-tuning'' while achieving a smaller model size and faster inference speed.

The automation of this field supported the development of novel Deep Learning (DL) [62] architectures, especially Convolutional Neural Networks (CNN) [61], that competed with previous state-of-the-art handcrafted models. Since the introduction of CNNs with LeNet [61] and the beginning of Deep Learning with AlexNet [60], most improvements in the field (e.g., deepening the architecture or adding residual connections) were driven by empiricism. NAS aims to put an end to this trial-and-error practice and bring a formal way to smoothen the progress in deep learning architecture design. Moreover, automatically discovering more efficient architectures is particularly relevant in the ecological transition context (i.e., green Deep Learning [119]). The role of manual feature engineering and model development has gradually decreased ever since.

Siamese networks are one of the most trending methods to achieve self-supervised visual representation learning (SSL). Since hand labeling is costly, SSL can play a crucial part by allowing deep learning to train on large unlabeled datasets. Meanwhile, Neural Architecture Search (NAS) is becoming increasingly important as a technique to discover novel deep learning architectures. However, early NAS methods based on reinforcement learning or evolutionary algorithms suffered from ludicrous computational and memory costs. In contrast, differentiable NAS, a gradient-based approach, has the advantage of being much more efficient and has thus retained most of the attention in the past few years. In this article, we present NASiam, a novel approach that uses for the first time differentiable NAS to improve the multilayer perceptron projector and predictor (encoder/predictor pair) architectures inside siamese-networks-based contrastive learning frameworks (e.g., SimCLR, SimSiam, and MoCo) while preserving the simplicity of previous baselines. We crafted a search space designed explicitly for multilayer perceptrons, inside which we explored several alternatives to the standard ReLU activation function. We show that these new architectures allow ResNet backbone convolutional models to learn strong representations efficiently. NASiam reaches competitive performance in both small-scale (i.e., CIFAR-10/CIFAR-100) and large-scale (i.e., ImageNet) image classification datasets while costing only a few GPU hours. We discuss the composition of the NAS-discovered architectures and emit hypotheses on why they manage to prevent collapsing behavior. Our code is available at https://github.com/aheuillet/NASiam.

Archaeologists are in dire need of automated object reconstruction methods. Fragments reassembly is close to puzzle problems, which may be solved by computer vision algorithms. As they are often beaten on most image related tasks by deep learning algorithms, we study a classification method that can solve jigsaw puzzles. In this paper, we focus on classifying the relative position: given a couple of fragments, we compute their local relation (e.g. on top). We propose several enhancements over the state of the art in this domain, which is outperformed by our method by 25\%. We propose an original dataset composed of pictures from the Metropolitan Museum of Art. We propose a greedy reconstruction method based on the predicted relative positions.