Canada

Knowledge distillation aims at transferring knowledge from the teacher model to the student one by aligning their distributions. 
Feature-level distillation often uses $\mathcal{L}_{2}$ distance or its variants as the loss function, based on the assumption that outputs follow normal distributions. 
This poses a significant challenge when distribution gaps are substantial since this loss function ignores the variance term. 
To address the problem, we propose to decompose the transfer objective into small parts and optimize it progressively. 
This process is inspired by diffusion models from which the noise distribution is mapped to the target distribution step by step.
However, directly employing diffusion models is impractical in the distillation scenario due to its heavy reverse process.
To overcome this challenge, we adopt the structural re-parameterization technique to generate multiple student features to approximate the teacher features sequentially. 
The multiple student features are combined linearly in inference time without extra cost.
We present extensive experiments performed on various transfer scenarios, such as CNN-to-CNN and Transformer-to-CNN, that validate the effectiveness of our approach.
Code is available at \url{https://github.com/yaoxufeng/KDiffusion}.

AAAI 2024

Progressively Knowledge Distillation via Re-parameterizing Diffusion Reverse Process

efficient transformer

diffusion

knowledge distillation

technical paper

We are pleased to announce the Thirty-Eighth AAAI Conference on Artificial Intelligence (AAAI-24), which will be held in Vancouver, British Columbia at the Vancouver Convention Centre – West Building from 20-27 February, 2024.

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-24 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.

We expect for AAAI-24 to be an in-person conference – one author of all accepted papers will be expected to present work in person unless there are exceptional circumstances that prevent this.<br><br><br><br>

In order to access the AAAI-24 event page you need to register [here](https://aaai.org/aaai-conference/registration/)

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. 

In the hydrology field, time series forecasting is crucial for efficient water resource management, improving flood and drought control and increasing the safety and quality of life for the general population. However, predicting long-term streamflow is a complex task due to the presence of extreme events. It requires the capture of long-range dependencies and the modeling of rare but important extreme values. Existing approaches often struggle to tackle these dual challenges simultaneously. In this paper, we specifically delve into these issues and propose Distance-weighted Auto-regularized Neural network (DAN), a novel extreme-adaptive model for long-range forecasting of stremflow enhanced by polar representation learning. DAN utilizes a distance-weighted multi-loss mechanism and stackable blocks to dynamically refine indicator sequences from exogenous data, while also being able to handle uni-variate time-series by employing Gaussian Mixture probability modeling to improve robustness to severe events. We also introduce Kruskal-Wallis sampling and gate control vectors to handle imbalanced extreme data. On four real-life hydrologic streamflow datasets, we demonstrate that DAN significantly outperforms both state-of-the-art hydrologic time series prediction methods and general methods designed for long-term time series prediction.

Learning from Polar Representation: An Extreme-Adaptive Model for Long-Term Time Series Forecasting

Non-native speakers with limited vocabulary often struggle to name specific objects despite being able to visualize them, e.g., people outside Australia searching for ‘numbats.’ Further, users may want to search for such elusive objects with difficult-to-sketch interactions, e.g., “numbat digging in the ground.” In such common but complex situations, users desire a search interface that accepts composite multimodal queries comprising hand-drawn sketches of “difficult-to-name but easy-to-draw” objects and text describing “difficult-to-sketch but easy-to-verbalize” object's attributes or interaction with the scene. This novel problem statement distinctly differs from the previously well-researched TBIR (text-based image retrieval) and SBIR (sketch-based image retrieval) problems. To study this under-explored task, we curate a dataset, CSTBIR (Composite Sketch+Text Based Image Retrieval), consisting of ~2M queries and 108K natural scene images. Further, as a solution to this problem, we propose a pretrained multimodal transformer-based baseline, STNet (Sketch+Text Network), that uses a hand-drawn sketch to localize relevant objects in the natural scene image, and encodes the text and image to perform image retrieval. In addition to contrastive learning, we propose multiple training objectives that improve the performance of our model. Extensive experiments show that our proposed method outperforms several state-of-the-art retrieval methods for text-only, sketch-only, and composite query modalities. We make the dataset and code available at: https://vl2g.github.io/projects/cstbir.

Composite Sketch+Text Queries for Retrieving Objects with Elusive Names and Complex Interactions

The softmax operator is a crucial component of large language models (LLMs), which have played a transformative role in computer research. Due to the centrality of the softmax function in the machine learning and convex optimization literature, past researchers have studied regression algorithms (one of the most fundamental methods in machine learning) on this function.

Much controversy has arisen around whether AI models simply regurgitate or copy data, especially if such data is copyrighted. Past researchers have developed methods to prevent generative machine learning models from producing copyrighted data. Our paper develops a novel method for preventing regression models from outputting copyrighted data (even if they are trained on such copyrighted data) by use of a simple, modified regression objective function, and uses this method to create a provably polynomial time regression algorithm for the softmax function that steers clear of copyright issues.

How to Protect Copyright Data in Optimization of Large Language Models?

Continuous sign language recognition (CSLR) aims to recognize gloss sequences from continuous sign videos. Recent works enhance the gloss representation consistency by mining correlations between visual and contextual modules within individual sentences. However, there still remain much richer correlations among glosses across different sentences. In this paper, we present a simple yet effective Cross-Sentence Gloss Consistency (CSGC), which enforces glosses belonging to a same category to be more consistent in representation than those belonging to different categories, across all training sentences. Specifically, in CSGC, a prototype is maintained for each gloss category and benefits the gloss discrimination in a contrastive way. Thanks to the well-distinguished gloss prototype, an auxiliary similarity classifier is devised to enhance the recognition clues, thus yielding more accurate results. Extensive experiments conducted on three CSLR datasets show that our proposed CSGC significantly boosts the performance of CSLR, surpassing existing state-of-the-art works by large margins (i.e., 1.6% on PHOENIX14, 2.4% on PHOENIX14-T, and 5.7% on CSL-Daily).

Cross-Sentence Gloss Consistency for Continuous Sign Language Recognition

In this paper, we first investigate a visual quality degradation problem observed in recent high-resolution virtual try-on approach. The tendency is empirically found that the textures of clothes are squeezed at the sleeve, as visualized in the upper row of Fig.1(a). A main reason for the issue arises from a gradient conflict between two popular losses, the Total Variation (TV) and adversarial losses. Specifically, the TV loss aims to disconnect boundaries between the sleeve and torso in a warped clothing mask, whereas the adversarial loss aims to combine between them. Such contrary objectives feedback the misaligned gradients to a cascaded appearance flow estimation, resulting in undesirable squeezing artifacts. To reduce this, we propose a Sequential Deformation (SD-VITON) that disentangles the appearance flow prediction layers into TV objective-dominant (TVOB) layers and a task-coexistence (TACO) layer. Specifically, we coarsely fit the clothes onto a human body via the TVOB layers, and then keep on refining via the TACO layer. In addition, the bottom row of Fig.1(a) shows a different type of squeezing artifacts around the waist. To address it, we further propose that we first warp the clothes into a tucked-out shirts style, and then partially erase the texture from the warped clothes without hurting the smoothness of the appearance flows. Experimental results show that our SD-VITON successfully resolves both types of artifacts and outperforms the baseline methods. Source code will be available at https://github.com/SHShim0513/SD-VITON.

Towards Squeezing-Averse Virtual Try-On via Sequential Deformation

Multi-view multi-label feature selection aims to select informative features where the data are collected from multiple sources with multiple interdependent class labels. For fully exploiting multi-view information, most prior works mainly focus on the common part in the ideal circumstance. However, the inconsistent part hidden in each view, including noises and specific elements, may affect the quality of mapping between labels and feature representations. Meanwhile, ignoring the specific part might lead to a suboptimal result, as each label is supposed to possess specific characteristics of its own. To deal with the double problems in multi-view multi-label feature selection, we propose a unified loss function which is a totally splitting structure for observed labels as hybrid labels that is, common labels, view-to-all specific labels and noisy labels, and the view-to-all specific labels further splits into several specific labels of each view. The proposed method simultaneously considers the consistency and complementarity of different views. Through exploring the feature weights of hybrid labels, the mapping relationships between labels and features can be established sequentially based on their attributes. Additionally, the interrelatedness among hybrid labels is also investigated and injected into the function. Specific to the specific labels of each view, we construct the novel regularization paradigm incorporating logic operations. Finally, the convergence of the result is proved after applying the multiplicative update rules. Experiments on six datasets demonstrate the effectiveness and superiority of our method compared with the state-of-the-art methods.

Double-Layer Hybrid-Label Identification Feature Selection for Multi-View Multi-Label Learning

Point-based interactive image segmentation can ease the burden of mask annotation in applications such as semantic segmentation and image editing. However, fully extracting the target mask with limited user inputs remains challenging. We introduce a novel method, Variance-Insensitive and Target-Preserving Mask Refinement to enhance segmentation quality with fewer user inputs. Regarding the last segmentation result as the initial mask, an iterative refinement process is commonly employed to continually enhance the initial mask. Nevertheless, conventional techniques suffer from sensitivity to the variance in the initial mask. To circumvent this problem, our proposed method incorporates a mask matching algorithm for ensuring consistent inferences from different types of initial masks. We also introduce a target-aware zooming algorithm to preserve object information during downsampling, balancing efficiency and accuracy. Experiments on GrabCut, Berkeley, SBD, and DAVIS datasets demonstrate our method's state-of-the-art performance in interactive image segmentation.

Variance-Insensitive and Target-Preserving Mask Refinement for Interactive Image Segmentation

Neural predictors have shown great potential in the evaluation process of neural architecture search (NAS). However, current predictor-based approaches overlook the fact that training a predictor necessitates a considerable number of trained neural networks as the labeled training set, which is costly to obtain. Therefore, the critical issue in utilizing predictors for NAS is to train a high-performance predictor using as few trained neural networks as possible. Although some methods attempt to address this problem through unsupervised learning, they often result in inaccurate predictions. We argue that the unsupervised tasks intended for the common graph data are too challenging for the graphs related to neural networks, causing unsupervised training to be susceptible to performance crashes. To address this issue, we propose a \textbf{C}urricu\textbf{L}um-guided \textbf{C}ontrastive \textbf{L}earning framework for neural \textbf{P}redictor (DCLP). Our method simplifies the contrastive task by designing a novel curriculum to enhance the stability of unlabeled training data distribution during contrastive training. Specifically, we propose a scheduler that ranks the training data according to the contrastive difficulty of each data and then inputs them to the contrastive learner in order. This approach concentrates the training data distribution and makes contrastive training more efficient. By using our method, the contrastive learner incrementally learns feature representations via unsupervised data on a smooth learning curve, avoiding performance crashes that may occur with excessively variable training data distributions. We experimentally demonstrate that DCLP has high accuracy and efficiency compared with existing predictors, and shows promising potential to discover superior architectures in various search spaces when combined with search strategies. Our code is available at: https://github.com/Zhengsh123/DCLP.

DCLP: Neural Architecture Predictor with Curriculum Contrastive Learning

Recently, Heterogeneous Federated Learning (HtFL) has attracted attention due to its ability to support heterogeneous models and data. To reduce the high communication cost of transmitting model parameters, a major challenge in HtFL, prototype-based HtFL methods are proposed to solely share class representatives, a.k.a, prototypes, among heterogeneous clients while maintaining the privacy of clients’ models. However, these prototypes are naively aggregated into global prototypes on the server using weighted averaging, resulting in suboptimal global knowledge which negatively impacts the performance of clients. To overcome this challenge, we introduce a novel HtFL approach called FedTGP, which leverages our Adaptive-margin-enhanced Contrastive Learning (ACL) to learn Trainable Global Prototypes (TGP) on the server. By incorporating ACL, our approach enhances prototype separability while preserving semantic meaning. Extensive experiments with twelve heterogeneous models demonstrate that our FedTGP surpasses state-of-the-art methods by up to 9.08% in accuracy while maintaining the communication and privacy advantages of prototype-based HtFL. Our code is available at https://github.com/TsingZ0/FedTGP.

FedTGP: Trainable Global Prototypes with Adaptive-Margin-Enhanced Contrastive Learning for Data and Model Heterogeneity in Federated Learning

Computational evaluations are crucial in modern problem-solving when we surpass theoretical algorithms or bounds. These experiments frequently take much work, and the sheer amount of needed resources makes it impossible to execute them on a single personal computer or laptop. Cluster schedulers allow for automatizing these tasks and scale to many computers. But, when we evaluate implementations of combinatorial algorithms, we depend on stable runtime results. Common approaches either limit parallelism or suffer from instable runtime measurements due to interference among jobs on modern hardware. The former is inefficient and not sustainable. The latter results in unreplicable experiments.
In this work, we address this issue and offer an acceptable balance between efficiency, software, hardware complexity, reliability, and replicability. We investigate effects towards replicability stability and illustrate how to efficiently use widely employed cluster resources for parallel evaluations. Furthermore, we present solutions which mitigate issues that emerge from the concurrent execution of benchmark jobs. Our experimental evaluation shows that – despite parallel execution – our approach reduces the runtime instability on the majority of instances to one second.

Progressively Knowledge Distillation via Re-parameterizing Diffusion Reverse Process

Downloads

Next from AAAI 2024

Learning from Polar Representation: An Extreme-Adaptive Model for Long-Term Time Series Forecasting

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Progressively Knowledge Distillation via Re-parameterizing Diffusion Reverse Process

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2024

Learning from Polar Representation: An Extreme-Adaptive Model for Long-Term Time Series Forecasting

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads