Browsing by Author "Malekipirbazari, Milad"

Now showing 1 - 12 of 12

Open Access
Feature selection using stochastic approximation with Barzilai and Borwein non-monotone gains
(Elsevier Ltd, 2021-08) Aksakallı, V.; Yenice, Z. D.; Malekipirbazari, Milad; Kargar, Kamyar
With recent emergence of machine learning problems with massive number of features, feature selection (FS) has become an ever-increasingly important tool to mitigate the effects of the so-called curse of dimensionality. FS aims to eliminate redundant and irrelevant features for models that are faster to train, easier to understand, and less prone to overfitting. This study presents a wrapper FS method based on Simultaneous Perturbation Stochastic Approximation (SPSA) with Barzilai and Borwein (BB) non-monotone gains within a pseudo-gradient descent framework wherein performance is measured via cross-validation. We illustrate that SPSA with BB gains (SPSA-BB) provides dramatic improvements in terms of the number of iterations for convergence with minimal degradation in cross-validated error performance over the current state-of-the art approach with monotone gains (SPSA-MON). In addition, SPSA-BB requires only one internal parameter and therefore it eliminates the need for careful fine-tuning of numerous other internal parameters as in SPSA-MON or comparable meta-heuristic FS methods such as genetic algorithms (GA). Our particular implementation includes gradient averaging as well as gain smoothing for better convergence properties. We present computational experiments on various public datasets with Nearest Neighbors and Naive Bayes classifiers as wrappers. We present comparisons of SPSA-BB against full set of features, SPSA-MON, as well as seven popular meta-heuristics based FS algorithms including GA and particle swarm optimization. Our results indicate that SPSA-BB converges to a good feature set in about 50 iterations on the average regardless of the number of features (whether a dozen or more than 1000 features) and its performance is quite competitive. SPSA-BB can be considered extremely fast for a wrapper method and therefore it stands as a high-performing new feature selection method that is also computationally feasible in practice.
Open Access
Formation of cluster mode particles (1–3 nm) in preschools
(Elsevier BV, 2021-11-23) Torkmahalleh, M. A.; Turganova, K.; Zhigulina, Z.; Madiyarova, T.; Adotey, E. K.; Malekipirbazari, Milad; Buonanno, G.; Stabile, L.
This study is the first study that reports the cluster particle (1–3 nm) formation (CPF) in two modern preschools located in Nur-Sultan city of Kazakhstan from October 28 to November 27, 2019. The average particle number concentration and mode diameter values during major CPF events in Preschool I and Preschool II were found to be 1.90 × 10^6 (SD 6.43 × 106) particles/cm3 and 1.60 (SD 0.85) nm, and 1.11 × 109 (SD 5.46 × 109) particles/cm3 and 2.16 (SD 1.47) nm, respectively. The ultraviolet PM concentration reached as high as 7 μg/m3 in one of the measurement days. The estimated emission rate in Preschool I for CPF events was 9.57 × 109 (SD 1.92 × 109) particles/min. For Preschool II, the emission rate was 7.25 × 109 (SD 12.4 × 109) particles/min. We identified primary cluster particles (CPs) emitted directly from the sources such as candle burning, and secondary CPs formed as a result of the oxidation of indoor VOCs or smoking VOCs. The secondary CPs are likely to be SOA. Indoor VOCs were mainly emitted during cleaning activities as well as during painting and gluing. Indoor VOCs are the controlling factors in the CPF events. Changes in the training and cleaning programs may result in significant reductions in the exposure of the children to CPs.
Open Access
Global air quality and COVID-19 pandemic: do we breathe cleaner air?
(Taiwan Association for Aerosol Research,Taiwan Qijiao Yanjiu Xuehui, 2021-02-08) Torkmahalleh, M. A.; Akhmetvaliyeva, Z.; Omran, A. D.; Omran, F. D.; Kazemitabar, M.; Naseri, M.; Motahareh, N.; Hamed, S.; Malekipirbazari, Milad; Adotey, E. K.; Soudabeh, G.; Neda, E.; Sabanov, S.; Alastuey, A.; Andrade, M. F.; Buonanno, G.; Carbone, S.; Cárdenas-Fuentes, D. E.; Cassee, F. R; Dai, Q.; Henríquez, A.; Hopke, P. K.; Keronen, P.; Khwaja, H. A.; Kim, J.; Kulmala, M.; Kumar, P.; Kushta, J.; Kuula, J.; Massagué, J.; Mitchell, T.; Mooibroek, D.; Morawska, L.; Niemi, J. V.; Ngagine, S. H.; Norman, M.; Oyama, B.; Oyola, P.; Öztürk, F.; Petäjä, T.; Querol, X.; Rashidi, Y.; Reyes, F.; Ross-Jones, M.; Salthammer, T.; Savvides, C.; Stabile, L.; Sjöberg, K.; Söderlund, K.; Raman, R. S.; Timonen, H.; Umezawa, M.; Viana, M.; Xie, S.
The global spread of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) has challenged most countries worldwide. It was quickly recognized that reduced activities (lockdowns) during the Coronavirus Disease of 2019 (COVID-19) pandemic produced major changes in air quality. Our objective was to assess the impacts of COVID-19 lockdowns on groundlevel PM2.5, NO2, and O3 concentrations on a global scale. We obtained data from 34 countries, 141 cities, and 458 air monitoring stations on 5 continents (few data from Africa). On a global average basis, a 34.0% reduction in NO2 concentration and a 15.0% reduction in PM2.5 were estimated during the strict lockdown period (until April 30, 2020). Global average O3 concentration increased by 86.0% during this same period. Individual country and continent-wise comparisons have been made between lockdown and business-as-usual periods. Universally, NO2 was the pollutant most affected by the COVID-19 pandemic. These effects were likely because its emissions were from sources that were typically restricted (i.e., surface traffic and non-essential industries) by the lockdowns and its short lifetime in the atmosphere. Our results indicate that lockdown measures and resulting reduced emissions reduced exposure to most harmful pollutants and could provide global-scale health benefits. However, the increased O3 may have substantially reduced those benefits and more detailed health assessments are required to accurately quantify the health gains. At the same, these restrictions were obtained at substantial economic costs and with other health issues (depression, suicide, spousal abuse, drug overdoses, etc.). Thus, any similar reductions in air pollution would need to be obtained without these extensive economic and other consequences produced by the imposed activity reductions.
Open Access
Human exposure to aerosol from indoor gas stove cooking and the resulting nervous system responses
(Wiley, 2022-01-17) Torkmahalleh, Mehdi Amouei; Naseri, Motahareh; Nurzhan, Sholpan; Gabdrashova, Raikhangul; Bekezhankyzy, Zhibek; Gimnkhan, Aidana; Malekipirbazari, Milad; Jouzizadeh, Mojtaba; Tabesh, Mahsa; Farrokhi, Hamta; Mehri-Dehnavi, Hossein; Khanbabaie, Reza; Sadeghi, Sahar; Khatir, Ali Alizadeh; Sabanov, Sergei; Buonanno, Giorgio; Hopke, Philip K.; Cassee, Flemming; Crape, Byron
Our knowledge of the effects of exposure to indoor ultrafine particles (sub-100 nm, #/cm3) on human brain activity is very limited. The effects of cooking ultrafine particles (UFP) on healthy adults were assessed using an electroencephalograph (EEGs) for brain response. Peak ultrafine particle concentrations were approximately 3 × 105 particle/cm3, and the average level was 1.64 × 105 particle/cm3. The average particle number emission rate (S) and the average number decay rate (a+k) for chicken frying in brain experiments were calculated to be 2.82 × 1012 (SD = 1.83 × 1012, R2 = 0.91, p = 0.0013) particles/min, 0.47 (SD = 0.30, R2 = 0.90, p < 0.0001) min−1, respectively. EEGs were recorded before and during cooking (14 min) and 30 min after the cooking sessions. The brain fast-wave band (beta) decreased during exposure, similar to people with neurodegenerative diseases. It subsequently increased to its pre-exposure condition for 70% of the study participants after 30 min. The brain slow-wave band to fast-wave band ratio (theta/beta ratio) increased during and after exposure, similar to observed behavior in early-stage Alzheimer's disease (AD) patients. The brain then tended to return to its normal condition within 30 min following the exposure. This study suggests that chronically exposed people to high concentrations of cooking aerosol might progress toward AD.
Open Access
The impact of frying aerosol on human brain activity
(Elsevier, 2019) Naseri, M.; Jouzizadeh, M.; Tabesh, M.; Malekipirbazari, Milad; Gabdrashova, R.; Nurzhan, S.; Farrokhi, H.; Khanbabaie, R.; Mehri-Dehnavi, H.; Bekezhankyzy, Z.; Gimnkhan, A.; Dareini, M.; Kurmangaliyeva, A.; Islam, N.; Crape, B.; Buonanno, G.; Cassee, F.; Torkmahalleh, M.
Knowledge on the impact of the exposure to indoor ultrafine particles (UFPs) on the human brain is restricted. Twelve non-atopic, non-smoking, and healthy adults (10 female and 7 male, in average 22 years old) were monitored for brain physiological responses via electroencephalographs (EEGs) during cooking. Frying ground beef meat in sunflower oil using electric stove without ventilation was conducted. UFPs, particulate matter (PM) (PM1, PM2.5, PM4, PM10), CO2, indoor temperature, RH, oil and meat temperatures were monitored continuously throughout the experiments. The UFP peak concentration was recorded to be approximately 2.0 × 105 particles/cm3. EEGs were recorded before exposure, at end of cooking when PM peak concentrations were observed, and 30 min after the end of the cooking session (post-exposure). Brain electrical activity statistically significantly changed during post-exposure compared to the before exposure, suggesting the translocation of UFPs to the brain, occurring solely in the frontal and temporal lobes of the brain. Study participants older than 25 were more susceptible to UFPs compared to those younger than 25. Also, the brain abnormality was mainly driven by male rather than female study participants. The brain slow-wave band (delta) decreased while the fast-wave band (Beta3) increased similar to the pattern found in the literature for the exposure to smoking fumes and diesel exhaust.
Open Access
The impact on heart rate and blood pressure following exposure to ultrafine particles from cooking using an electric stove
(Elsevier, 2020-08-02) Gabdrashova, R.; Nurzhan, S.; Naseri, M.; Bekezhankyzy, Z.; Gimnkhan, A.; Malekipirbazari, Milad; Tabesh, M.; Khanbabaie, R.; Crape, B.; Buonanno, G.; Hopke, P. K.; Torkmahalleh, A. A.; Torkmahalleh, M. A.; Aleya, L.
Cooking is a major source of indoor particulate matter (PM), especially ultrafine particles (UFPs). Long-term exposure to fine and ultrafine particles (UFPs) has been associated with adverse human health effects. Toxicological studies have demonstrated that exposure to PM2.5 (particles with aerodynamic diameter smaller than 2.5 μm) may result in increased blood pressure (BP). Some clinical studies have shown that acute exposure to PM2.5 causes changes in systolic (SBP) and diastolic blood pressure (DBP), depending on the source of particles. Studies assessing the effect of exposure to cooking PM on BP and heart rate (HR) using electric or gas stoves are not well represented in the literature. The aim of this investigation was to perform controlled studies to quantify the exposure of 50 healthy volunteer participants to fine and ultrafine particles emitted from a low-emissions recipe for frying ground beef on an electric stove. The BP and heart rate (HR) of the volunteers were monitored during exposure and after the exposure (2 h post-exposure). Maximum UFP and PM2.5 concentrations were 6.5 × 104 particles/cm3 and 0.017 mg/m3, respectively. Exposure to UFPs from frying was associated with statistically significant increases in the SBP. The lack of food and drink during the 2 h post-cooking period was also associated with a statistically significant reduction in SBP. No statistically significant changes in DBP were observed. Physiological factors, including heat stress over the stove, movements and anxiety, could be responsible for an elevation in HR at the early stages of the experiments with a subsequent drop in HR after 90 min post-cooking, when study participants were relaxed in a living room.
Open Access
In Press, Corrected Proof: Formation of cluster mode particles (1–3 nm) in preschools
(Elsevier, 2021-11-23) Torkmahalleh, M. A.; Turganova, K.; Zhigulina, Z.; Madiyarova, T.; Adotey, E. K.; Malekipirbazari, Milad; Milad, G.; Stabile, L.
This study is the first study that reports the cluster particle (1–3 nm) formation (CPF) in two modern preschools located in Nur-Sultan city of Kazakhstan from October 28 to November 27, 2019. The average particle number concentration and mode diameter values during major CPF events in Preschool I and Preschool II were found to be 1.90 × 106 (SD 6.43 × 106) particles/cm3 and 1.60 (SD 0.85) nm, and 1.11 × 109 (SD 5.46 × 109) particles/cm3 and 2.16 (SD 1.47) nm, respectively. The ultraviolet PM concentration reached as high as 7 μg/m3 in one of the measurement days. The estimated emission rate in Preschool I for CPF events was 9.57 × 109 (SD 1.92 × 109) particles/min. For Preschool II, the emission rate was 7.25 × 109 (SD 12.4 × 109) particles/min. We identified primary cluster particles (CPs) emitted directly from the sources such as candle burning, and secondary CPs formed as a result of the oxidation of indoor VOCs or smoking VOCs. The secondary CPs are likely to be SOA. Indoor VOCs were mainly emitted during cleaning activities as well as during painting and gluing. Indoor VOCs are the controlling factors in the CPF events. Changes in the training and cleaning programs may result in significant reductions in the exposure of the children to CPs.
Embargo
Index policy for multiarmed bandit problem with dynamic risk measures
(Elsevier BV, 2023-08-06) Malekipirbazari, Milad; Çavus, Özlem
The multiarmed bandit problem (MAB) is a classic problem in which a finite amount of resources must be allocated among competing choices with the aim of identifying a policy that maximizes the expected total reward. MAB has a wide range of applications including clinical trials, portfolio design, tuning parameters, internet advertisement, auction mechanisms, adaptive routing in networks, and project management. The classical MAB makes the strong assumption that the decision maker is risk-neutral and indifferent to the variability of the outcome. However, in many real life applications, these assumptions are not met and decision makers are risk-averse. Motivated to resolve this, we study risk-averse control of the multiarmed bandit problem in regard to the concept of dynamic coherent risk measures to determine a policy with the best risk-adjusted total discounted return. In respect of this specific setting, we present a theoretical analysis based on Whittle’s retirement problem and propose a priority-index policy that reduces to the Gittins index when the level of risk-aversion converges to zero. We generalize the restart formulation of the Gittins index to effectively compute these risk-averse allocation indices. Numerical results exhibit the excellent performance of this heuristic approach for two well-known coherent risk measures of first-order mean-semideviation and mean-AVaR. Our experimental studies suggest that there is no guarantee that an index-based optimal policy exists for the risk-averse problem. Nonetheless, our risk-averse allocation indices can achieve optimal or near-optimal policies which in some instances are easier to interpret compared to the exact optimal policy.
Open Access
Modeling the rheological properties of highly nano-filled polymers
(SAGE Publications Ltd, 2017) Kourki, Hajir; Mortezaei, Mehrzad; Famili, Mohammad Hossein Navid; Malekipirbazari, Milad
Organic and inorganic materials are usually added to polymers in order to achieve some benefits such as reducing the product cost, as well as achieving higher modulus and strength. Addition of these materials would change polymers’ behavior. Adding nano-materials to polymers on the other hand is a new challenge in the field of polymer composites where previous studies were unable to achieve good correlation with nano-composites at higher particle volume fractions. In this research, Yamamoto network theory is developed to investigate the behavior of highly nano-filled systems. For this purpose, five different types of sub-chain and two types of junctions are considered and the effect of particle size, concentration, and the model parameters in association with the behavior of the junctions are studied. Moreover, some experiments are performed on polystyrene filled with nano-silica at different particle size and concentration values in frequency mod in the linear region. At last, we compared the results of our final model with the experiments in order to evaluate its accuracy, which confirmed a very good agreement. © 2016, © The Author(s) 2016.
Open Access
Performance comparison of feature selection and extraction methods with random instance selection
(Elsevier Ltd, 2021-10-01) Malekipirbazari, Milad; Aksakallı, V.; Shafqat, W.; Eberhard, A.
In pattern recognition, irrelevant and redundant features together with a large number of noisy instances in the underlying dataset decrease performance of trained models and make the training process considerably slower, if not practically infeasible. In order to combat this so-called curse of dimensionality, one option is to resort to feature selection (FS) methods designed to select the features that contribute the most to the performance of the model, and one other option is to utilize feature extraction (FE) methods that map the original feature space into a new space with lower dimensionality. These two methods together are called feature reduction (FR) methods. On the other hand, deploying an FR method on a dataset with massive number of instances can become a major challenge, from both memory and run time perspectives, due to the complex numerical computations involved in the process. The research question we consider in this study is rather a simple, yet novel one: do these FR methods really need the whole set of instances (WSI) available for the best performance, or can we achieve similar performance levels with selecting a much smaller random subset of WSI prior to deploying an FR method? In this work, we provide empirical evidence based on comprehensive computational experiments that the answer to this critical research question is in the affirmative. Specifically, with simple random instance selection followed by FR, the amount of data needed for training a classifier can be drastically reduced with minimal impact on classification performance. We also provide recommendations on which FS/ FE method to use in conjunction with which classifier.
Open Access
Risk-averse allocation indices for multiarmed bandit problem
(IEEE, 2021-01-25) Malekipirbazari, Milad; Çavuş, Özlem
In classical multiarmed bandit problem, the aim is to find a policy maximizing the expected total reward, implicitly assuming that the decision-maker is risk-neutral. On the other hand, the decision-makers are risk-averse in some real-life applications. In this article, we design a new setting based on the concept of dynamic risk measures where the aim is to find a policy with the best risk-adjusted total discounted outcome. We provide a theoretical analysis of multiarmed bandit problem with respect to this novel setting and propose a priority-index heuristic which gives risk-averse allocation indices having a structure similar to Gittins index. Although an optimal policy is shown not always to have index-based form, empirical results express the excellence of this heuristic and show that with risk-averse allocation indices we can achieve optimal or near-optimal interpretable policies.
Open Access
Risk-averse multi-armed bandit problem
(2021-08) Malekipirbazari, Milad
In classical multi-armed bandit problem, the aim is to ﬁnd a policy maximizing the expected total reward, implicitly assuming that the decision maker is risk-neutral. On the other hand, the decision makers are risk-averse in some real life applications. In this study, we design a new setting for the classical multi-armed bandit problem (MAB) based on the concept of dynamic risk measures, where the aim is to ﬁnd a policy with the best risk adjusted total discounted outcome. We provide theoretical analysis of MAB with respect to this novel setting, and propose two diﬀerent priority-index heuristics giving risk-averse allocation indices with structures similar to Gittins index. The ﬁrst proposed heuristic is based on Lagrangian duality and the indices are expressed as the Lagrangian multiplier corresponding to the activation constraint. In the second part, we present a theoretical analysis based on Whittle’s retirement problem and propose a gener-alized version of restart-in-state formulation of the Gittins index to compute the proposed risk-averse allocation indices. Finally, as a practical application of the proposed methods, we focus on optimal design of clinical trials and we apply our risk-averse MAB approach to perform risk-averse treatment allocation based on a Bayesian Bernoulli model. We evaluate the performance of our approach against other allocation rules, including ﬁxed randomization.