Seyed Mohammad Jafar Jalali's research while affiliated with Deakin University and other places

... Reliability is an important measure defined in the context of RSs, which relies on the concept that how much the produced recommendations are reliable for users. It is shown that the more reliable items are recommended, the more relevant the recommendations will be to the user's preferences [19,[65][66][67]. Therefore, considering the reliability of predictions in RSs can significantly improve the accuracy of recommendations, which results in increasing user's satisfaction. ...

... This way, they do not have to go through the complex process of extracting features. CNNs have been applied successfully in medical image classification [12], detection [13], segmentation [14,15], and other image processing tasks [16]. Many studies on brain tumor segmentation have used CNN-based methods, which have greatly improved the speed and accuracy of diagnosing brain tumors. ...

... A significant gap exists between the prediction results obtained by these methods and ERL. [56] PM2.5 forecasting Q-learning 2021 Li et al. [82] PM2.5 forecasting Q-learning 2021 Chao Chen and Hui Liu [83] wind speed prediction deep Q-Network 2021 Jalali et al. [84] wind power forecasting Q-learning 2022 Tan et al. [31] PM2.5 prediction Sarsa 2022 Qin et al. [85] unit commitment problem deep Q-Network 2022 Sogabe et al. [86] smart energy optimization and risk evaluation Q-learning 2022 Sharma et al. [52] estimating reference evapotranspiration Q-learning 2022 He et al. [87] wind farm control deep deterministic policy gradient 2022 Jalali et al. [88] solar irradiance forecasting Q-learning 2022 Shi Yin and Hui Liu [61] wind power prediction Q-learning 2023 Yu et al. [29] wind power prediction deep deterministic policy gradient [89], who used an online algorithm. The performance of the ERL method has been verified on simulation platforms [90], and the experimental results ...

... A largely hidden layer with more parameters usually improves the prediction accuracy but dramatically increases the number of weights. And a small hidden layer does not propagate all input features well, resulting in suboptimal results [41]. To make up for both shortcomings and solve the parameter redundancy and overfitting caused by a fully connected layer, we design an IFC layer. ...

... Wu et al. employed the swarm intelligence algorithm to optimize the welding sequence optimization and FNNs; this algorithm eliminated premature convergence and generated the best solution [22]. Raziani et al. combined a modified whale optimization algorithm based on a nonlinear function with FNNs to resolve medical classification problems; this method had faster operation efficiency and better evaluation indexes [23]. Dong et al. designed an efficient and reliable training algorithm to solve FNNs; this algorithm utilized flexibility and stability to obtain a better objective value [24]. ...

... Zhao et al. [32] and Lei et al. [33] performed attribute reduction based on rough set, and then combined fuzzy Bandelet neural network and deep learning to predict the life of impeller and building energy consumption, respectively. Khodayar et al. [34] proposed a novel deep generative neural network for wind speed forecasting by combining with rough set theory. Sheikhoushaghi et al. [35] designed a rough neural network to forecast oil production rate. ...

... With the consideration of temporal and spatial variation in scene time series, Yu et al. [29] developed a wind power forecasting method based on the convolutional neural network (CNN). To optimize the hyper-parameters of the CNN for wind power forecasting, Jalali et al. [30] proposed an improved version of the grey wolf optimization algorithm. In reference [31], spatio-temporal correlation characteristics were represented using correlation coefficients and the Shapley value method, and they were employed in the power forecasting based on CNN and long-short term memory (LSTM) networks. ...

... The FCL's output is then passed through the last layer, which is usually a softmax layer. The softmax function is a common choice for the output layer of a classification network, as it produces a probability distribution over the possible classes [50]. The softmax function takes a vector of arbitrary real-valued scores and produces a vector of probabilities that sums to one. ...

... Moreover, the accurate diagnosis of COVID-19 infection is highly dependent on the expertise of radiologists, which poses difficulties for some due to its recent emergence and similarity to other lung diseases, such as pneumonia (Luz et al., 2022). Thus, such issues necessitate the development of an automated decision support system to improve the diagnosis's efficiency, accuracy, and speed, which can be achieved using machine learning and statistical algorithms (Jalali et al., 2022). ...

... In other words, these models choose random values for the weights and biases of ANNs that lead to a high-level uncertainty in the training process. Moreover, falling into local optima can negatively affect the convergence speed of the training process [7][8][9][10][11]. Accordingly, choosing appropriate initial values of weights and biases is a critical issue to find global optima in the training process of MLP networks. ...