Learning with Linear Mixed Model for Group Recommendation Systems

Gao, Baode; Zhan, Guangpeng; Wang, Hanzhang; Wang, Yiming; Zhu, Shufang

doi:10.1145/3318299.3318342

Cited by 8 publications

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Since then, pruning techniques gained popularity, searching for effective methods to prune parameters (Zhang et al, 2021;Frankle & Carbin, 2018;Lee et al, 2019). For assessing pruning and quantization algorithms, Gao et al (2019); Isik et al (2022) provided a rate-distortion theory framework, showing that entropy reduction during training is beneficial, as low-entropy models are more amenable to compression (Oktay et al, 2019;Baskin et al, 2019). This study continues these guidelines, using rate-distortion theory to analyze rotation-invariant solutions and further provides enhancements that find the optimal solution efficiently.…”

Section: Related Workmentioning

confidence: 99%

Rotation Invariant Quantization for Model Compression

Kampeas¹,

Nahshan²,

Kremer³

et al. 2023

Preprint

View full text Add to dashboard Cite

Post-training Neural Network (NN) model compression is an attractive approach for deploying large, memory-consuming models on devices with limited memory resources. In this study, we investigate the rate-distortion tradeoff for NN model compression. First, we suggest a Rotation-Invariant Quantization (RIQ) technique that utilizes a single parameter to quantize the entire NN model, yielding a different rate at each layer, i.e., mixed-precision quantization. Then, we prove that our rotation-invariant approach is optimal in terms of compression. We rigorously evaluate RIQ and demonstrate its capabilities on various models and tasks. For example, RIQ facilitates ×19.4 and ×52.9 compression ratios on pretrained VGG dense and pruned models, respectively, with < 0.4% accuracy degradation. Code: https://github.com/ehaleva/RIQ.

show abstract

Section: Related Workmentioning

confidence: 99%

Rotation Invariant Quantization for Model Compression

Kampeas¹,

Nahshan²,

Kremer³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…However, the inclusion of more participants with more repeated measures could enable an expansion of the statistical model to also account for participant-specific random slopes. That would enable modeling participant-specific sensitivity toward contextual predictors (Harrison et al 2018;Gao et al 2019).…”

Section: Limitationsmentioning

confidence: 99%

Measuring and modeling context-dependent preferences for hearing aid settings

Pasta

Petersen

Jensen

et al. 2022

User Model User-Adap Inter

View full text Add to dashboard Cite

Despite having individual perceptual preferences toward sounds, hearing aid users often end up with default hearing aid settings that have no contextual awareness. However, the introduction of smartphone-connected hearing aids has enabled a rethinking of hearing aids as user-adaptive systems considering both individual and contextual differences. In this study, we aimed to investigate the feasibility of such context-aware system for providing hearing aid users with a number of relevant hearing aid settings to choose from. During normal real-world hearing aid usage, we applied a smartphone-based method for capturing participants’ listening experience and audiological preference for different intervention levels of three audiological parameters (Noise Reduction, Brightness, Soft Gain). Concurrently, we collected contextual data as both self-reports (listening environment and listening intention) and continuous data logging of the acoustic environment (sound pressure level, signal-to-noise ratio). First, we found that having access to different intervention levels of the Brightness and Soft Gain parameters affected listening satisfaction. Second, for all three audiological parameters, the perceived usefulness of having access to different intervention levels was significantly modulated by context. Third, contextual data improved the prediction of both explicit and implicit intervention level preferences. Our findings highlight that context has a significant impact on hearing aid preferences across participants and that contextual data logging can help reduce the space of potential interventions in a user-adaptive system so that the most useful and preferred settings can be offered. Moreover, the proposed mixed-effects model is suitable for capturing predictions on an individual level and could also be expanded to predictions on a group level by including relevant user features.

show abstract

“…Such difficulty is referred to as the cold start problem. The introduction of linear mixed model (LMM) can help us with predicting the preference of these new users if we have some information other than the watching history [10]. Thinking about the favorite movies of people of different ages, we may have the intuition that people of different ages are likely to show distinct tastes or preferences.…”

Section: Introductionmentioning

confidence: 99%

“…It should be noted that the linear mixed model has long been used for knowledge/science discovery in breeding [12], and now widely used in ecology [13], genetics [14], [15], and genome-wide association study [16], [17]. However, the application of LMM in recommendation systems was examined just recently by sparse publications, for example [10], [11], [18]. The underlying mathematics foundation and theoretical details were inadequate or even omitted in all such recommendation literatures.…”

Section: Introductionmentioning

confidence: 99%

Knowledge Discovery and Recommendation With Linear Mixed Model

et al. 2020

Self Cite

View full text Add to dashboard Cite

We give a concise tutorial on knowledge discovery with linear mixed model in movie recommendation. The versatility of mixed effects model is well explained. Commonly used methods for parameter estimation, confidence interval estimate and evaluation criteria for model selection are briefly reviewed. Mixed effects models produce sound inference based on a series of rigorous analysis. In particular, we analyze millions of movie rating data with LME4 R package and find solid evidences for a general social behavior: the young tend to be more censorious than senior people when evaluating the same object. Such a social behavior phenomenon can be used in recommender systems and business data analysis. INDEX TERMS Knowledge discovery in database (KDD), linear mixed-effects model (LMM), recommender system (RS), R software.

show abstract

Learning with Linear Mixed Model for Group Recommendation Systems

Cited by 8 publications

References 11 publications

Rotation Invariant Quantization for Model Compression

Rotation Invariant Quantization for Model Compression

Measuring and modeling context-dependent preferences for hearing aid settings

Knowledge Discovery and Recommendation With Linear Mixed Model

Contact Info

Product

Resources

About