Modern deep neural networks, particularly recent large language models, come with massive model sizes that require significant computational and storage resources. To enable the deployment of modern models on resource-constrained environments and to accelerate inference time, researchers have increasingly explored pruning techniques as a popular research direction in neural network compression. More than three thousand pruning papers have been published from 2020 to 2024. However, there is a dearth of up-to-date comprehensive review papers on pruning. To address this issue, in this survey, we provide a comprehensive review of existing research works on deep neural network pruning in a taxonomy of 1) universal/specific speedup, 2) when to prune, 3) how to prune, and 4) fusion of pruning and other compression techniques. We then provide a thorough comparative analysis of eight pairs of contrast settings for pruning (e.g., unstructured/structured, one-shot/iterative, data-free/data-driven, initialized/pre-trained weights, etc.) and explore several emerging topics, including pruning for large language models, vision transformers, diffusion models, and large multimodal models, post-training pruning, and different levels of supervision for pruning to shed light on the commonalities and differences of existing methods and lay the foundation for further method development. Finally, we provide some valuable recommendations on selecting pruning methods and prospect several promising research directions for neural network pruning. To facilitate future research on deep neural network pruning, we summarize broad pruning applications (e.g., adversarial robustness, natural language understanding, etc.) and build a curated collection of datasets, networks, and evaluations on different applications. We maintain a repository on https://github.com/hrcheng1066/awesome-pruning that serves as a comprehensive resource for neural network pruning papers and corresponding open-source codes. We will keep updating this repository to include the latest advancements in the field.

Battiti's mutual information feature selector (MIFS) and its variant algorithms are used for many classification applications. Since they ignore feature synergy, MIFS and its variants may cause a big bias when features are combined to cooperate together. Besides, MIFS and its variants estimate feature redundancy regardless of the corresponding classification task. In this paper, we propose an automated greedy feature selection algorithm called conditional mutual information-based feature selection (CMIFS). Based on the link between interaction information and conditional mutual information, CMIFS takes account of both redundancy and synergy interactions of features and identifies discriminative features. In addition, CMIFS combines feature redundancy evaluation with classification tasks. It can decrease the probability of mistaking important features as redundant features in searching process. The experimental results show that CMIFS can achieve higher best-classification-accuracy than MIFS and its variants, with the same or less (nearly 50%) number of features.

A challenge of channel pruning is designing efficient and effective criteria to select channels to prune. A widely used criterion is minimal performance degeneration, e.g., loss changes before and after pruning being the smallest. To accurately evaluate the truth performance degeneration requires retraining the survived weights to convergence, which is prohibitively slow. Hence existing pruning methods settle to use previous weights (without retraining) to evaluate the performance degeneration. However, we observe that the loss changes differ significantly with and without retraining. It motivates us to develop a technique to evaluate true loss changes without retraining, using which to select channels to prune with more reliability and confidence. We first derive a closed-form estimator of the true loss change per mask change, using influence functions without retraining. Influence function is a classic technique from robust statistics that reveals the impacts of a training sample on the model's prediction and is repurposed by us to assess impacts on true loss changes. We then show how to assess the importance of all channels simultaneously and develop a novel global channel pruning algorithm accordingly. We conduct extensive experiments to verify the effectiveness of the proposed algorithm, which significantly outperforms the competing channel pruning methods on both image classification and object detection tasks. To the best of our knowledge, we are the first that shows evaluating true loss changes or pruning without retraining is possible. This finding will open up opportunities for a series of new paradigms to emerge that differ from existing pruning methods.

The poor performance and the lack of manual labeled corpus are two main problems in the task of protein-protein interaction extraction. A novel hybrid method is proposed. Based on the individual characteristics of machine learning and pattern learning, this method utilizes learned patterns from pattern learning to generate pattern features by performing sequence alignment. The pattern features and word features are incorporated into the input feature set of machine learning algorithms. The semi-supervised method based on k-nearest neighbours classifier is also proposed to train the hybrid method from unlabeled data automatically. Experimental results show the improved performance over the baseline methods with the hybrid model and the efficieny of the semi-supervised method for the lack of labeled data.

Facial age estimation is of interest due to its potential to be applied in many real-life situations. However, recent age estimation efforts do not consider juveniles. Consequently, we introduce a juvenile age detection scheme called LaGMO, which focuses on the juvenile aging cues of facial shape and appearance. LaGMO is a combination of facial landmark points and Term Frequency Inverse Gravity Moment (TF-IGM). Inspired by the formation of words from morphemes, we obtained facial appearance features comprising facial shape and wrinkle texture and represented them as terms that described the age of the face. By leveraging the implicit ordinal relationship between the frequencies of the terms in the face, TF-IGM was used to compute the weights of the terms. From these weights, we built a matrix that corresponds to the possibilities of the face belonging to the age. Next, we reduced the reference matrix according to the juvenile age range (0–17 years) and avoided the exhaustive search through the entire training set. LaGMO detects the age by the projection of an unlabeled face image onto the reference matrix; the value of the projection depicts the higher probability of the image belonging to the age. With Mean Absolute Error (MAE) of 89% on the Face and Gesture Recognition Research Network (FG-NET) dataset, our proposal demonstrated superior performance in juvenile age estimation.