We introduce AlphaGrad, a memory-efficient, conditionally stateless optimizer addressing the memory overhead and hyperparameter complexity of adaptive methods like Adam. AlphaGrad enforces scale invariance via tensor-wise L2 gradient normalization followed by a smooth hyperbolic tangent transformation, $g' = \tanh(\alpha \cdot \tilde{g})$, controlled by a single steepness parameter $\alpha$. Our contributions include: (1) the AlphaGrad algorithm formulation; (2) a formal non-convex convergence analysis guaranteeing stationarity; (3) extensive empirical evaluation on diverse RL benchmarks (DQN, TD3, PPO). Compared to Adam, AlphaGrad demonstrates a highly context-dependent performance profile. While exhibiting instability in off-policy DQN, it provides enhanced training stability with competitive results in TD3 (requiring careful $\alpha$ tuning) and achieves substantially superior performance in on-policy PPO. These results underscore the critical importance of empirical $\alpha$ selection, revealing strong interactions between the optimizer's dynamics and the underlying RL algorithm. AlphaGrad presents a compelling alternative optimizer for memory-constrained scenarios and shows significant promise for on-policy learning regimes where its stability and efficiency advantages can be particularly impactful.

We give new results for problems in computational and statistical machine learning using tools from high-dimensional geometry and probability. We break up our treatment into two parts. In Part I, we focus on computational considerations in optimization. Specifically, we give new algorithms for approximating convex polytopes in a stream, sparsification and robust least squares regression, and dueling optimization. In Part II, we give new statistical guarantees for data science problems. In particular, we formulate a new model in which we analyze statistical properties of backdoor data poisoning attacks, and we study the robustness of graph clustering algorithms to ``helpful'' misspecification.

Learning music theory not only has practical benefits for musicians to write, perform, understand, and express music better, but also for both non-musicians to improve critical thinking, math analytical skills, and music appreciation. However, current external tools applicable for learning music theory through writing when human instruction is unavailable are either limited in feedback, lacking a written modality, or assuming already strong familiarity of music theory concepts. In this paper, we describe Maestoso, an educational tool for novice learners to learn music theory through sketching practice of quizzed music structures. Maestoso first automatically recognizes students' sketched input of quizzed concepts, then relies on existing sketch and gesture recognition techniques to automatically recognize the input, and finally generates instructor-emulated feedback. From our evaluations, we demonstrate that Maestoso performs reasonably well on recognizing music structure elements and that novice students can comfortably grasp introductory music theory in a single session.

Kanji script writing is a skill that is often introduced to novice Japanese foreign language students for achieving Japanese writing mastery, but often poses difficulties to students with primarily English fluency due to their its vast differences with written English. Instructors often introduce various pedagogical methods -- such as visual structure and written techniques -- to assist students in kanji study, but may lack availability providing direct feedback on students' writing outside of class. Current educational applications are also limited due to lacking richer instructor-emulated feedback. We introduce Kanji Workbook, a writing-based intelligent tutoring system for students to receive intelligent assessment that emulates human instructor feedback. Our interface not only leverages students' computing devices for allowing them to learn, practice, and review the writing of prompted characters from their course's kanji script lessons, but also provides a diverse set of writing assessment metrics -- derived from instructor interviews and classroom observation insights -- through intelligent scoring and visual animations. We deployed our interface onto novice- and intermediate-level university courses over an entire academic year, and observed that interface users on average achieved higher course grades than their peers and also reacted positively to our interface's various features.

Fish Feeding Intensity Assessment (FFIA) is crucial in industrial aquaculture management. Recent multi-modal approaches have shown promise in improving FFIA robustness and efficiency. However, these methods face significant challenges when adapting to new fish species or environments due to catastrophic forgetting and the lack of suitable datasets. To address these limitations, we first introduce AV-CIL-FFIA, a new dataset comprising 81,932 labelled audio-visual clips capturing feeding intensities across six different fish species in real aquaculture environments. Then, we pioneer audio-visual class incremental learning (CIL) for FFIA and demonstrate through benchmarking on AV-CIL-FFIA that it significantly outperforms single-modality methods. Existing CIL methods rely heavily on historical data. Exemplar-based approaches store raw samples, creating storage challenges, while exemplar-free methods avoid data storage but struggle to distinguish subtle feeding intensity variations across different fish species. To overcome these limitations, we introduce HAIL-FFIA, a novel audio-visual class-incremental learning framework that bridges this gap with a prototype-based approach that achieves exemplar-free efficiency while preserving essential knowledge through compact feature representations. Specifically, HAIL-FFIA employs hierarchical representation learning with a dual-path knowledge preservation mechanism that separates general intensity knowledge from fish-specific characteristics. Additionally, it features a dynamic modality balancing system that adaptively adjusts the importance of audio versus visual information based on feeding behaviour stages. Experimental results show that HAIL-FFIA is superior to SOTA methods on AV-CIL-FFIA, achieving higher accuracy with lower storage needs while effectively mitigating catastrophic forgetting in incremental fish species learning.

The training of artificial neural networks is heavily dependent on the careful selection of an appropriate loss function. While commonly used loss functions, such as cross-entropy and mean squared error (MSE), generally suffice for a broad range of tasks, challenges often emerge due to limitations in data quality or inefficiencies within the learning process. In such circumstances, the integration of supplementary terms into the loss function can serve to address these challenges, enhancing both model performance and robustness. Two prominent techniques, loss regularization and contrastive learning, have been identified as effective strategies for augmenting the capacity of loss functions in artificial neural networks. Knowledge tracing is a compelling area of research that leverages predictive artificial intelligence to facilitate the automation of personalized and efficient educational experiences for students. In this paper, we provide a comprehensive review of the deep learning-based knowledge tracing (DKT) algorithms trained using advanced loss functions and discuss their improvements over prior techniques. We discuss contrastive knowledge tracing algorithms, such as Bi-CLKT, CL4KT, SP-CLKT, CoSKT, and prediction-consistent DKT, providing performance benchmarks and insights into real-world deployment challenges. The survey concludes with future research directions, including hybrid loss strategies and context-aware modeling.

Detecting collaborative and problem-solving behaviours from digital traces to interpret students' collaborative problem solving (CPS) competency is a long-term goal in the Artificial Intelligence in Education (AIEd) field. Although multimodal data and advanced models are argued to have the potential to detect complex CPS behaviours, empirical evidence on their value remains limited with some contrasting evidence. In this study, we investigated the potential of multimodal data to improve model performance in diagnosing 78 secondary school students' CPS subskills and indicators in authentic educational settings. In particular, text embeddings from verbal data and acoustic embed-dings from audio data were used in a multimodal classification model for CPS diagnosis. Both unimodal and multimodal transformer-based models outperformed traditional models in detecting CPS classes. Although the inclusion of multimodality did not improve the performance of traditional unimodal models, its integration into transformer-based models demonstrated improved performance for diagnosing social-cognitive CPS classes compared to unimodal transformer-based models. Based on the results, the paper argues that multimodality and the selection of a particular modelling technique should not be taken for granted to achieve the best performance in the automated detection of every CPS subskill and indicator. Rather, their value is limited to certain types of CPS indicators, affected by the complexity of the labels, and dependent on the composition of indicators in the dataset. We conclude the paper by discussing the required nuance when considering the value of LLMs and multimodality in automated CPS diagnosis, highlighting the need for human-AI complementarity, and proposing the exploration of relevant model architectures and techniques to improve CPS diagnosis in authentic educational contexts.

While large language models (LLMs) have been widely applied to code generation, they struggle with generating entire deep learning projects, which are characterized by complex structures, longer functions, and stronger reliance on domain knowledge than general-purpose code. An open-domain LLM often lacks coherent contextual guidance and domain expertise for specific projects, making it challenging to produce complete code that fully meets user requirements. In this paper, we propose a novel planning-guided code generation method, DLCodeGen, tailored for generating deep learning projects. DLCodeGen predicts a structured solution plan, offering global guidance for LLMs to generate the project. The generated plan is then leveraged to retrieve semantically analogous code samples and subsequently abstract a code template. To effectively integrate these multiple retrieval-augmented techniques, a comparative learning mechanism is designed to generate the final code. We validate the effectiveness of our approach on a dataset we build for deep learning code generation. Experimental results demonstrate that DLCodeGen outperforms other baselines, achieving improvements of 9.7% in CodeBLEU and 3.6% in human evaluation metrics.

Enhancing computational efficiency and reducing deployment costs for large language models (LLMs) have become critical challenges in various resource-constrained scenarios. In this work, we present DistilQwen2.5, a family of distilled, lightweight LLMs derived from the public Qwen2.5 models. These distilled models exhibit enhanced instruction-following capabilities compared to the original models based on a series of distillation techniques that incorporate knowledge from much larger LLMs. In our industrial practice, we first leverage powerful proprietary LLMs with varying capacities as multi-agent teachers to select, rewrite, and refine instruction-response pairs that are more suitable for student LLMs to learn. After standard fine-tuning, we further leverage a computationally efficient model fusion approach that enables student models to progressively integrate fine-grained hidden knowledge from their teachers. Experimental evaluations demonstrate that the distilled models possess significantly stronger capabilities than their original checkpoints. Additionally, we present use cases to illustrate the applications of our framework in real-world scenarios. To facilitate practical use, we have released all the DistilQwen2.5 models to the open-source community.

The process of creating educational materials is both time-consuming and demanding for educators. This research explores the potential of Large Language Models (LLMs) to streamline this task by automating the generation of extended reading materials and relevant course suggestions. Using the TED-Ed Dig Deeper sections as an initial exploration, we investigate how supplementary articles can be enriched with contextual knowledge and connected to additional learning resources. Our method begins by generating extended articles from video transcripts, leveraging LLMs to include historical insights, cultural examples, and illustrative anecdotes. A recommendation system employing semantic similarity ranking identifies related courses, followed by an LLM-based refinement process to enhance relevance. The final articles are tailored to seamlessly integrate these recommendations, ensuring they remain cohesive and informative. Experimental evaluations demonstrate that our model produces high-quality content and accurate course suggestions, assessed through metrics such as Hit Rate, semantic similarity, and coherence. Our experimental analysis highlight the nuanced differences between the generated and existing materials, underscoring the model's capacity to offer more engaging and accessible learning experiences. This study showcases how LLMs can bridge the gap between core content and supplementary learning, providing students with additional recommended resources while also assisting teachers in designing educational materials.