GPU Accelerated Feature Engineering and Training for Recommender Systems

Schifferer, Benedikt; Titericz, Gilberto; Deotte, Chris; Henkel, Christof; Onodera, Kazuki; Liu, Jiwei; Tunguz, Bojan; Oldridge, Even; Moreira, Gabriel de Souza Pereira; Erdem, A.

doi:10.1145/3415959.3415996

Cited by 17 publications

(6 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Together with GBMs, DNNs usually achieve the best results in machine learning competitions [ 37 , 38 ]. DNNs were used for Retention Time (RT) prediction in [ 21 ], where a DNN with 4 layers and regularization was proposed.…”

Section: Methodsmentioning

confidence: 99%

Probabilistic metabolite annotation using retention time prediction and meta-learned projections

et al. 2022

View full text Add to dashboard Cite

Retention time information is used for metabolite annotation in metabolomic experiments. But its usefulness is hindered by the availability of experimental retention time data in metabolomic databases, and by the lack of reproducibility between different chromatographic methods. Accurate prediction of retention time for a given chromatographic method would be a valuable support for metabolite annotation. We have trained state-of-the-art machine learning regressors using the 80, 038 experimental retention times from the METLIN Small Molecule Retention Tim (SMRT) dataset. The models included deep neural networks, deep kernel learning, several gradient boosting models, and a blending approach. 5, 666 molecular descriptors and 2, 214 fingerprints (MACCS166, Extended Connectivity, and Path Fingerprints fingerprints) were generated with the alvaDesc software. The models were trained using only the descriptors, only the fingerprints, and both types of features simultaneously. Bayesian hyperparameter search was used for parameter tuning. To avoid data-leakage when reporting the performance metrics, nested cross-validation was employed. The best results were obtained by a heavily regularized deep neural network trained with cosine annealing warm restarts and stochastic weight averaging, achieving a mean and median absolute errors of $$39.2 \pm 1.2\; s$$ 39.2 ± 1.2 s and $$17.2 \pm 0.9\;s$$ 17.2 ± 0.9 s , respectively. To the best of our knowledge, these are the most accurate predictions published up to date over the SMRT dataset. To project retention times between chromatographic methods, a novel Bayesian meta-learning approach that can learn from just a few molecules is proposed. By applying this projection between the deep neural network retention time predictions and a given chromatographic method, our approach can be integrated into a metabolite annotation workflow to obtain z-scores for the candidate annotations. To this end, it is enough that just as few as 10 molecules of a given experiment have been identified (probably by using pure metabolite standards). The use of z-scores permits considering the uncertainty in the projection when ranking candidates, and not only the accuracy. In this scenario, our results show that in 68% of the cases the correct molecule was among the top three candidates filtered by mass and ranked according to z-scores. This shows the usefulness of this information to support metabolite annotation. Python code is available on GitHub at https://github.com/constantino-garcia/cmmrt.

show abstract

Section: Methodsmentioning

confidence: 99%

Probabilistic metabolite annotation using retention time prediction and meta-learned projections

et al. 2022

View full text Add to dashboard Cite

show abstract

“…XGBoost (Chen et al, 2015), LightGBM (Ke et al, 2017) andCatBoost (Dorogush et al, 2018) are considered for gradient boosting models. Since the originally proposed XGBoost uses LE, Kfold TE (Ayria, 2020) is applied to XGBoost following Schifferer et al (2020). Considered CTR prediction models are as follows: FM (Rendle, 2010), FFM (Juan et al, 2016), AFM (Xiao et al, 2017), DCN , MLP, NFM , Wide & Deep (Cheng et al, 2016), DeepFM (Guo et al, 2017), xDeepFM (Lian et al, 2018), PNN (Qu et al, 2018), AutoInt (Song et al, 2019), and AFN (Cheng et al, 2020).…”

Section: Methodsmentioning

confidence: 99%

“…Meanwhile, TE changes the categorical feature into an informative number by calculating the mean of target values with each categorical feature (Micci-Barreca, 2001). However, TE causes overfitting by giving excessive information on each categorical feature (Schifferer et al, 2020).…”

Section: Efficient Click-through Rate Predictionmentioning

confidence: 99%

See 1 more Smart Citation

Efficient Click-Through Rate Prediction for Developing Countries via Tabular Learning

Yi¹,

Chang²

2021

Preprint

View full text Add to dashboard Cite

Despite the rapid growth of online advertisement in developing countries, existing highly over-parameterized Click-Through Rate (CTR) prediction models are difficult to be deployed due to the limited computing resources. In this paper, by bridging the relationship between CTR prediction task and tabular learning, we present that tabular learning models are more efficient and effective in CTR prediction than over-parameterized CTR prediction models. Extensive experiments on eight public CTR prediction datasets show that tabular learning models outperform twelve state-of-the-art CTR prediction models. Furthermore, compared to over-parameterized CTR prediction models, tabular learning models can be fast trained without expensive computing resources including high-performance GPUs. Finally, through an A/B test on an actual online application, we show that tabular learning models improve not only offline performance but also the CTR of real users.

show abstract

“…To address this limitation, we plan to integrate highperformance computing, such as graphical processor units [40] and supercomputing [41], as future work.…”

Section: Funding Limitation and Future Perspectivesmentioning

confidence: 99%

Toward a Cognitive-Inspired Hashtag Recommendation for Twitter Data Analysis

Djenouri

Belhadi

Srivastava

et al. 2022

IEEE Trans. Comput. Soc. Syst.

View full text Add to dashboard Cite

This research investigates hashtag suggestions in a heterogeneous and huge social network, as well as a cognitive-based deep learning solution based on distributed knowledge graphs. Community detection is first performed to find the connected communities in a vast and heterogeneous social network. The knowledge graph is subsequently generated for each discovered community, with an emphasis on expressing the semantic relationships among the Twitter platform's user communities. Each community is trained with the embedded deep learning model. To recommend hashtags for the new user in the social network, the correlation between the tweets of such user and the knowledge graph of each community is explored to set the relevant communities of such user. The models of the relevant communities are used to infer the hashtags of the tweets of such users. We conducted extensive testing to demonstrate the usefulness of our methods on a variety of tweet collections. Experimental results show that the proposed approach is more efficient than the baseline approaches in terms of both runtime and accuracy.

show abstract

GPU Accelerated Feature Engineering and Training for Recommender Systems

Cited by 17 publications

References 7 publications

Probabilistic metabolite annotation using retention time prediction and meta-learned projections

Probabilistic metabolite annotation using retention time prediction and meta-learned projections

Efficient Click-Through Rate Prediction for Developing Countries via Tabular Learning

Toward a Cognitive-Inspired Hashtag Recommendation for Twitter Data Analysis

Contact Info

Product

Resources

About