Cold Start Research Articles

Hybrid CNN-based Recommendation System

Recommendation systems are now being used to address the problem of excess information in several sectors such as entertainment, social networking, and e-commerce. Although conventional methods to recommendation systems have achieved significant success in providing item suggestions, they still face many challenges, including the cold start problem and data sparsity. Numerous recommendation models have been created in order to address these difficulties. Nevertheless, including user or item-specific information has the potential to enhance the performance of recommendations. The ConvFM model is a novel convolutional neural network architecture that combines the capabilities of deep learning for feature extraction with the effectiveness of factorization machines for recommendation tasks. The present work introduces a novel hybrid deep factorization machine (FM) model, referred to as ConvFM. The ConvFM model use a combination of feature extraction and convolutional neural networks (CNNs) to extract features from both individuals and things, namely movies. Following this, the proposed model employs a methodology known as factorization machines, which use the FM algorithm. The focus of the CNN is on the extraction of features, which has resulted in a notable improvement in performance. In order to enhance the accuracy of predictions and address the challenges posed by sparsity, the proposed model incorporates both the extracted attributes and explicit interactions between items and users. This paper presents the experimental procedures and outcomes conducted on the Movie Lens dataset. In this discussion, we engage in an analysis of our research outcomes followed by provide recommendations for further action.

Effect analysis on hydrocarbon adsorption enhancement of different zeolites in cold start of gasoline engine based on Monte Carlo method

50%–80% of the hydrocarbons in gasoline engine emissions come from during cold starts. Zeolites are often used to adsorb HCs during gasoline engine cold-starting. In order to investigate the effect of adsorption enhancement of different molecular sieves on HCs during gasoline engine cold-starting. In this paper, the adsorption properties of Beta, MOR and FER zeolites at different temperatures were simulated based on Monte Carlo method using propylene and toluene as probe molecules. The results show that all three molecular sieves exhibit good adsorption properties at low temperatures. The adsorption capacity decreased with increasing temperature. The molecular sieves tend to adsorb molecules with diameters similar to the pore size of zeolite. In addition, the three-dimensional zeolites Beta and FER showed better adsorption properties than the one-dimensional zeolites. Based on Monte Carlo method, the adsorption properties of Beta, MOR and FER molecular sieves were investigated at different temperatures under multi-components using toluene, isopentane, propylene, ethylene, methane and water as adsorbents. The results showed that there was a competitive adsorption among the components due to the limited adsorption sites in the molecular sieve channels. Large molecules such as toluene and isopentane occupied most of the adsorption sites and hindered the adsorption of small molecules.

A Novel Cross-Domain Recommendation with Evolution Learning

In this “info-plosion” era, recommendation systems (or recommenders) play a significant role in finding interesting items in the surge of online digital activities and e-commerce. Several techniques have been widely applied for recommendation systems, but the cold-start and sparsity problems remain a major challenge. The cold-start problem occurs when generating recommendations for new users and items without sufficient information. Sparsity refers to the problem of having a large amount of users and items but with few transactions or interactions. In this article, a novel cross-domain recommendation model, Cross-Domain Evolution Learning Recommendation (abbreviated as CD-ELR), is developed to communicate the information from different domains in order to tackle the cold-start and sparsity issues by integrating matrix factorization and recurrent neural network. We introduce an evolutionary concept to describe the preference variation of users over time. Furthermore, several optimization methods are developed for combining the domain features for precision recommendation. Experimental results show that CD-ELR outperforms existing state-of-the-art recommendation baselines. Finally, we conduct experiments on several real-world datasets to demonstrate the practicability of the proposed CD-ELR.

MapReduce based on serverless platforms

This study intends to investigate the application of the MapReduce (MR) framework based on serverless computing in big data processing. By combining the MapReduce model with serverless computing, efficient data processing is achieved. In this framework, the phases of Map task execution, reduce task execution, etc. are accomplished through stateless serverless functions, and data storage is realized with the help of cloud storage platforms (e.g., OSS). In this paper, the author introduces the basic theory of MR, the basic theory of serverless computing, describes the framework implementation process, and discusses the role of OSS in distributed computing. The outcomes of the trial indicate the average execution time of the framework for a WordCount task on 100 pieces of TOEFL English reading data is 6.81 seconds. The discussion analyzes the frameworks advantages (high elasticity, resource utilization) and disadvantages (cold start latency, unsuitable for long time tasks). Future research directions encompass performance optimization, long-time task processing, state management, etc. In summary, this study provides valuable insights for the practical application of serverless computing in big data processing.

Experimental visualization of water/ice phase distribution at cold start for practical-sized polymer electrolyte fuel cells

The automotive industry aims to ensure the cold-start capability of polymer electrolyte fuel cells (PEFCs) for developing fuel cell electric vehicles that can be driven in cold climates. Water and ice behavior is a key factor in maintaining this capability. Previously reported methods for visualizing water and/or ice are limited to small-sized PEFCs ( < 50 cm2), while fuel cell electric vehicles are equipped with larger PEFCs. Here, we developed a system using a pulsed spallation neutron beam to visualize water distribution and identify water/ice phases in practical-sized PEFCs at a cold start. The results show direct evidence of a stepwise freezing behavior inside the PEFC. The produced water initially accumulated at the center of the PEFC and then froze, followed by PEFC shutdown as freezing progressed. This study can serve as a reference to guide the development of cold-start protocols, cell design, and materials for next-generation fuel cell electric vehicles.

BCE4ZSR: Bi-encoder empowered by teacher cross-encoder for zero-shot cold-start news recommendation

In the realm of news recommendations, the persistent challenge of the cold-start problem continues to impede progress. Existing approaches rely heavily on information exchange between news articles and users to personalize news recommendations and have struggled to adapt to users and news articles without historical interaction data. In this paper, we introduce BCE4ZSR, a novel framework that leverages a rarely utilized zero-shot approach to effectively tackle the cold-start problem in news recommendations. The proposed approach consists of two main steps: First, we generate embeddings for inference with a sentence transformer (a bi-encoder). In the second step, a fine-tuned transformer model is augmented during the training phase to distil the bi-encoder with knowledge from the cross-encoder model using a student–teacher framework. As a result of this synergy, the cross-encoder serves as a teacher, imparting its knowledge to the bi-encoder. The proposed technique can be applied to any neural news recommender system and is empirically evaluated in both cold-start and regular user-news interaction situations. Experiments on real-world benchmark datasets (MIND-small and MIND-large) indicated that BCE4ZSR outperformed the baseline methods in terms of nDCG@k, AUC, and MRR. Specifically, AUC improvement of 1.5%–6% & 2.2%–7.93%; MRR improved by 1%–13.85% & 0.9%–14.72%; nDCG@5 improved by 4.9%–18.79% & 2.1%–16.44%; and nDCG@10 improved by 1.7%–15.18% & 1.9%–14.24% in the cold start and warm user scenarios respectively, proving the superiority of our model compared to baseline methods.

Nonlinear controller for SEPIC with single variable to tune

This work proposes a systematic approach to designing a novel integral sliding mode controller (ISMC) for a single-ended primary-inductor converter (SEPIC) with only one tunable parameter where the upper and lower bounds are derived. The designed surface results in minimal chattering behaviour at the output voltage as well as at the duty cycle and allows for operating the SEPIC at a fixed switching frequency. The proposed controller can withstand up to a 70% variation in the input voltage and 100% variation on the load side, in addition to superior performance for a cold start. The proposed controller and the corresponding mathematical formulation were simulated in a Simulink environment and experimentally tested via a scaled prototype. The proposed controller performance is also compared to Type II and integral Linear-Quadratic Regulators (LQR).

Study on transient thermal state of steel piston in diesel engine under cold start condition

The thermal load issue of pistons remains a topic of active research, especially the transient thermal state of steel pistons in diesel engines under cold start conditions has not been studied in depth. In this paper, a new steel piston power assembly was designed and developed on the basis of a small-bore diesel engine. A real-time temperature test system for the steel piston temperature field was developed to address the problem of the inability to obtain real-time temperature data. The test system was applied to the steel piston to investigate the transient variations in temperature field, thermal stress, and thermal deformation at different starting temperatures. The results show that the temperature field of steel pistons exhibits exponential growth in the early stage of cold start and then tends to stabilize after about 44 s. The lower the cold start temperature and the closer the monitoring points are to the high-temperature combustion gas, the faster the rate of temperature rise. The thermal stress of the steel piston initially increases rapidly, then decreases, and subsequently increases again until reaching a stable level. The lower the starting temperature, the longer the thermal stress inflection point of the monitoring point will be. The variation of thermal deformation of the steel piston follows the same trend as that of the temperature field.

PRICELESS: Privacy enhanced AI‐driven scalable framework for IoT applications in serverless edge computing environments

AbstractServerless edge computing has emerged as a new paradigm that integrates the serverless and edge computing. By bringing processing power closer to the edge of the network, it provides advantages such as low latency by quickly processing data for time‐sensitive Internet of Things (IoT) applications. Additionally, serverless edge computing also brings inherent problems of edge and serverless computing such as cold start, security and privacy that are still waiting to be solved. In this paper, we propose a new Blockchain‐based AI‐driven scalable framework called PRICELESS, to offer security and privacy in serverless edge computing environments while performing cold start prediction. In PRICELESS framework, we used deep reinforcement learning for the cold start latency prediction. For experiments, a cold start dataset is created using a heart disease risk‐based IoT application and deployed using Google Cloud Functions. Experimental results show the additional delay that the blockchain module brings to cold start latency and its impact on cold start prediction performance. Additionally, the performance of PRICELESS is compared with the current state‐of‐the‐art method based on energy cost, computation time and cold start prediction. Specifically, it has been observed that PRICELESS causes 19 ms of external latency, 358.2 watts for training, and 3.6 watts for prediction operations, resulting in additional energy consumption at the expense of security and privacy.

Improvement of the recommendation system based on the multi-armed bandit algorithm

In order to effectively solve common problems of the recommendation system, such as the cold start problem and dynamic data modeling problem, the multi-armed bandit (MAB) algorithm, the collaborative filtering (CF) algorithm, and the user information feedback are applied by researchers to update the recommendation model online and in time. In other words, the cold start problem of the recommendation system is transformed into an issue of exploration and utilization. The MAB algorithm is used, user features are introduced as content, and the synergy between users is further considered. In this paper, the author studies the improvement of the recommendation system based on the multi-armed bandit algorithm. The Liner Upper Confidence Bound (LinUCB), Collaborative Filtering Bandits (COFIBA), and Context-Aware clustering of Bandits (CAB) algorithms are analyzed. It is found that the MAB algorithm can get a good maximum total revenue regardless of the content value after going through the cold start stage. In the case of a particularly large amount of content, the CAB algorithm achieves the greatest effect.

Study on the forming process of the cylinder wall fuel film and the piston top land ablation mechanism of a medium speed marine diesel engine under cold start condition

The study presents an explanation of the piston top land ablation mechanism in a specific type of marine medium-speed diesel engine, utilizing multi-dimensional Computational Fluid Dynamics (CFD) software. At present, there is a lack of research on the process analysis of fuel droplets attached to the cylinder wall at diesel engine cold start condition and the mechanism of piston top land ablation. The study results indicate that during low-speed cold start condition, a multitude of diesel droplets impact the wall, causing the combustion area concentrated near the wall of the combustion chamber. The last stage fuel spray droplets, driven by the vortex in the cylinder, adhere to the cylinder wall before evaporation during the post-combustion stage. After multi-cycle operation of the diesel engine under cold start condition, the mass of fuel film adhering to the cylinder wall and the cumulative fuel mass on the top of the gas ring increase. As the engine getting warm, higher cylinder wall temperature leads to rapid evaporation of the fuel film on the cylinder wall and the gas ring. Deflagration happens when the fuel vapor burns under high temperature and high-pressure condition, and subsequent pressure oscillation destroys the thermal boundary layer of the piston top land. Piston material ablation happens under high temperature and high-pressure shock.

Design of A Portable Solid-Fuel Rocket Stove

The depletion of fossil fuels is a hot topic of discussion these days. This is due to the increasing use of fossil fuels in human life. In terms of the increase in fuel and gas prices, we realise that the energy consumption that is increasing from year to year is not balanced with the availability of energy sources, which affects households that use fuel in the cooking process. So a design was carried out to make an ergonomic solid fuel stove with the aim of not using fuel anymore in the cooking process, in this design, the stove is designed to be more flexible and get a portable biomass rocket stove design by considering social aspects, performance, local resources, economy, environmentally friendly. Biomass fuels are solids, liquids or gases produced from organic matter.. Biomass fuels used in solid fuel stoves such as firewood, charcoal, briquettes and pellets. Each biomass fuel has different properties and characteristics. The designed portable rocket stove consists of several main parts namely combustion chamber, chimney, air duct, pan spot and cassing, using stainless steel material. The test results include combustion chamber and water temperature during the process, thermal efficiency, combustion rate and specific fuel consumption under cold start and hot start conditions. After hot start, it is found that the use of a chimney on a portable stove designed using coconut shell fuel is more effective than without a chimney. Shorter boiling time of 1 minute, greater combustion rate of 0.13 grams/minute, higher thermal efficiency of 2% and lower specific fuel consumption of 9.11 grams/litre of water when the stove uses a chimney.

December 2023. Cold and wintry start, then very mild and wet with stormy episodes

December 2023. Cold and wintry start, then very mild and wet with stormy episodes

Experimental investigation of atomization characteristics at low temperature environments during diesel engine cold start

Exploring the atomization characteristics of fuel in low-temperature environments is crucial for improving the cold start performance of diesel engines. In this work, based on LIEF-PIV laser testing technology, a diesel engine cold start fuel atomization experimental system was built, which could simulate the internal environment of a diesel engine's cylinder during cold start. The concentration and velocity distributions of free-jet spray and wall-impinging spray at low temperature were measured experimentally. The results showed that the decrease in ambient temperature had a significant impact on the rate of evaporation of fuel droplets, which was mainly reflected in two aspects: the total evaporation of the fuel and the uniformity of the distribution of vapor phase. The decrease in fuel temperature led to a decrease in the conversion efficiency of vapor/liquid phases of the fuel. When the fuel temperature was −20 °C, the distribution range of vapor equivalence ratio near the wall area was 0.11–0.59, which is significantly lower than the other two conditions. In addition, the increase in environmental pressure resulted in a certain decrease in the maximum flow velocity and turbulent flow energy of the mixture, with an average decrease of 0.32 m/s in the maximum flow velocity at the three fuel temperatures of 0 °C, −20 °C, and −40 °C.

Matrix Factorization and Cosine Similarity Based Recommendation System For Cold Start Problem in E-Commerce Industries

Matrix Factorization and Cosine Similarity Based Recommendation System For Cold Start Problem in E-Commerce Industries

LR-MPIBS: A LoRa-Based Maritime Position-Indicating Beacon System

Human marine activities are becoming increasingly frequent. The adverse marine environment has led to an increase in man overboard incidents, resulting in significant losses of life and property. After a drowning accident, the accurate location information of the drowning victim can help improve the success rate of rescue. In this paper, we explore a LoRa-based Maritime Position-Indicating Beacon System (LR-MPIBS). A low-power drowning detection circuit is designed in LR-MPIBS to detect drowning accidents in a timely manner after a person falls into the water. The instantaneous high current of the LoRa RF can lower the supply voltage and cause other modules to work abnormally. A fast current transient response circuit is proposed to solve the problem. LR-MPIBS includes a power ripple suppression circuit that can reduce the measurement errors and operational abnormalities caused by power ripple interference. We explore the impedance matching law of LoRa RF circuits through simulation experiments to improve the quality of LoRa communication. A data processing algorithm for personnel drift trajectory is proposed to alleviate the challenges caused by the raw positioning data with large deviations and high communication cost. The experimental results show that LR-MPIBS can automatically start and actively alarm within 3 s after a person falls into the water. The positioning cold start time is less than 50 s. The performance of communication distance is more than 5 km. The endurance of LR-MPIBS is 25 h (with a 30 s communication cycle).

-30°C cold start optimization of PEMFC based on a data-driven surrogate model and multi-objective optimization algorithm

Cold start is a critical operating scenario for the proton exchange membrane fuel cell (PEMFC), particularly in the field of transportation. Under sub-freezing temperatures, the water inside the cell will freeze and obstruct gas flow paths as well as cover catalyst reaction sites, resulting in a failed startup. This study proposes an optimization method for the -30°C cold start of PEMFC based on a data-driven surrogate model to improve cold start performance and reduce irreversible damage to the cell. A validated PEMFC cold start mechanism model is utilized as the basis for developing an extreme learning machine (ELM) based data-driven surrogate model, which is trained using data collected from the mechanism model and has higher computational efficiency compared with the original model. In addition, the NSGA-II multi-objective optimization algorithm is employed to optimize the current loading strategies and operating parameters using the surrogate model as fitness function. The objectives are to enhance the minimum voltage and reduce startup duration time. Moreover, experimental validation confirms the effectiveness of the proposed method. The test results demonstrate that a cold start from -30°C is achieved within 97 s, with the minimum voltage reaching 0.44 V. Notably, there is a reduction in startup time by 26 s and an increase in the minimum voltage by 0.06 V compared to the base case. This study establishes a foundation for researchers to adjust operating settings during cold start based on diverse applications and requirements.

Block-level Image Service for the Cloud

Businesses increasingly need agile and elastic computing infrastructure to respond quickly to real-world situations. By offering efficient process-based virtualization and a layered image system, containers are designed to enable agile and elastic application deployment. However, creating or updating large container clusters is still slow due to the image downloading and unpacking process. In this article, we present DADI Image Service (DADI), a block-level image service for increased agility and elasticity in deploying applications. DADI replaces the waterfall model of starting containers (downloading image, unpacking image, starting container) with fine-grained on-demand transfer of remote images, realizing instant start of containers. To accelerate the cold start of containers, DADI designs a pull-based prefetching mechanism that allows a host to read necessary image data beforehand at the granularity of image layers. We design a peer-to-peer–based decentralized image sharing architecture to balance traffic among all the participating hosts and propose a pull-push collaborative prefetching mechanism to accelerate cold start. DADI efficiently supports various kinds of runtimes including cgroups, QEMU, and so on, further realizing “build once, run anywhere.” DADI has been deployed at scale in the production environment of Alibaba, serving one of the world’s largest ecommerce platforms. Performance results show that DADI can cold start 10,000 containers on 1,000 hosts within 4 s.

Prebaking runtime environments to improve the FaaS cold start latency

Function-as-service (FaaS) platforms promise a simpler programming model for cloud computing, given that providers take care of the overall resource management while the developers can concentrate only on writing their applications in the scope of a function, not having to care about installing or scaling resources. As FaaS users are billed based on the execution of the functions, platform providers have a natural incentive not to keep idle resources running at the platform’s expense. However, this strategy may lead to the cold start issue, in which the execution of a function is delayed because there are no ready resources to host the execution. Despite the time to provision the computational resources, starting the runtime environment that manages the function execution can take hundreds of milliseconds to seconds, a prohibitive non-deterministic overhead to latency-critical applications. This work describes a practical technique, Prebaking, to reduce the function start-up time based on restoring snapshots of previously executed processes. The foundation of the Prebaking technique is that deciding when to create a snapshot of a function is essential. For example, based on a prototype we developed using the CRIU checkpoint/restore Linux tool, the performance improvement on the start-up of a properly snapshotted function is up to 25 times. Fortunately, generating a good function snapshot is easy: One only needs to checkpoint a warm function, that is, a function that has processed a request. To show the feasibility of using the Prebaking technique, we discussed integrating it into OpenFaaS, an open-source FaaS platform, including how to generate and restore function snapshots. We also evaluated our prototype by running a comprehensive set of experiments that compare the function start-up duration and cold start latency against the standard fork-exec procedure. We analyze the JVM, CPython, and Node.js runtimes. The results indicate that the technique can improve the function replica start-up time even for no-warmed functions: we improved the function replica start-up time from 3.5 times for a “do-nothing” function running in Node.js up to 12 times when considering an Image Resizer function running in CPython. Finally, we reinforce the practicability of the proposed technique by comparing our proposal to another snapshot-based technique called SEUSS. That evaluation shows that Prebaking provides better cold start latency even for slightly complex functions like the Markdown Renderer.

Real-Time Movie Recommendation: Integrating Persona-Based User Modeling with NMF and Deep Neural Networks

The proliferation of uncategorized information on the Internet has intensified the need for effective recommender systems. Recommender systems have evolved from content-based filtering to collaborative filtering and, most recently, to deep learning-based and hybrid models. However, they often face challenges such as high computational costs, reduced reliability, and the Cold Start problem. We introduce a persona-based user modeling approach for real-time movie recommendations. Our system employs Non-negative Matrix Factorization (NMF) and Deep Learning algorithms to manage complex and sparse data types and to mitigate the Cold Start issue. Experimental results, based on criteria involving 50 topics and 35 personas, indicate a significant performance gain. Specifically, with 500 users, the precision@K for NMF was 86.01%, and for the Deep Neural Network (DNN), it was 92.67%. Tested with 900 users, the precision@K for NMF increased to 97.04%, and for DNN, it was 95.55%. These results represent an approximate 10% and 5% improvement in performance, respectively. The system not only delivers fast and accurate recommendations but also reduces computational overhead by updating the model only when user personas change. The generated user personas can be adapted for other recommendation services or large-scale data mining.