ABAnjanava Biswas

Anjanava Biswas

*This is an algorithmically generated author record based on publicly available information.

Anjanava Biswas has a total of 10 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Electric Vehicles Charging Station Load Forecasting Integration With Renewable Energy Using Novel Deep EfficientBiLSTMNet

    by Vineet Dhanawat, Varun Shinde, Rachid Alami, Adnan Akhunzada, Zaid Bin Faheem, Anjanava Biswas

    The exponential increase in the adoption of Electric Vehicles (EVs) presents significant problems to the stability of the power grid. Therefore, it is crucial to accurately anticipate the demand for EV Charging Station (CS) to address this issue. To improve forecasts and identify CS load variables, existing studies are based on load profiling, which may be difficult to obtain for commercial EV charging stations. This paper proposes an efficient deep BiLSTMNet model to solve and mitigate these problems. Energy consumption and storage at four charging stations in California are analyzed. To guarantee accuracy and uniformity, the data is preprocessed by addressing missing values and ensuring consistency. A hybrid feature selection technique integrates the Boruta algorithm and SHAP (SHapley Additive exPlanations) values to ensure robust feature selection. The EfficientBiLSTMNet model, which integrates the EfficientNet and BiLSTM layers, is trained on the preprocessed datasets. The model's hyperparameters are optimized using an Enhanced Firefly Algorithm (EFA). The model performs a time series analysis to identify daily, weekly, monthly, and seasonal patterns in EV charging demand. The integration of renewable energy sources-specifically solar and wind generation-into the EV charging infrastructure is thoroughly examined in this study, not merely as input features but as key factors influencing the stability of charging demand at various stations. Their temporal patterns and environmental dependencies are leveraged to enhance forecasting accuracy and ensure grid-aware demand management across charging stations. The proposed model's performance is evaluated using metrics such as R-squared, Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE). Simulation results demonstrate the effectiveness of the proposed model, with an average R-squared value of 0.9, MAE of 2.15 kW, and RMSE of 2.75 kW across the four stations. The EfficientBiLSTMNet model shows superior predictive accuracy compared to traditional models, highlighting the importance of comprehensive feature selection and engineering in forecasting EV charging demand. This study provides a robust framework for predicting EV charging demand, integrating renewable energy sources to enhance the stability and sustainability of the power grid amidst the increasing penetration of EVs.


  • Open Access

    Corrections to “Copy-Move Forgery Detection Technique Using Graph Convolutional Networks Feature Extraction”

    by Varun Shinde, Vineet Dhanawat, Ahmad Almogren, Anjanava Biswas, Muhammad Bilal, Rizwan Ali Naqvi, Ateeq Ur Rehman

    Presents corrections to the paper, (Corrections to “Copy-Move Forgery Detection Technique Using Graph Convolutional Networks Feature Extraction”).

  • The Long Context Conundrum: Challenges and Innovations in Scaling LLM Memory

    by Anjanava Biswas

    In an era where AI assistants are expected to digest entire novels, multi-document dossiers, or hours of conversation, the demand for longer-context large language models (LLMs) is soaring. Top AI companies are racing to expand context windows: OpenAI’s GPT-4 introduced 32,000-token versions, Anthropic’s Claude leapt from a 9K to 100K token window (and later to 200K), and startups like AI21 have unveiled models with 256K-token capacity [1]. Google DeepMind is reportedly aiming even higher, with hints of million-token contexts in its upcoming Gemini model [1]. These numbers promise that an LLM could ingest hundreds of pages or hours of transcribed audio in one go. But with great context comes great computation – extending an LLM’s memory isn’t as simple as turning a dial. Each additional token in context exacts a steep cost in processing time and memory usage. The long-context conundrum is emerging as one of the toughest challenges in scaling AI capabilities.

  • Open Access

    Blockchain Enabled Federated Learning for Detection of Malicious Internet of Things Nodes

    by Rachid Alami, Anjanava Biswas, Varun Shinde, Ahmad Almogren, Ateeq Ur Rehman, Tahseen Shaikh

    The Internet of Things (IoTs) networks are evolving day by day as they have been used in almost every field of life in the last few decades. The reason for the increasing trend of IoT networks is due to the increasing population of the world. However, these networks are vulnerable to the presence of malicious nodes, which compromise the efficiency of the decision-making process in the IoT network. Many machine learning and artificial intelligence techniques are proposed to solve this issue. Centralized learning is performed in these conventional machine learning techniques due to which the privacy of the network is compromised. Therefore, internal users are not encouraged to share their sensitive information in the network and external users do not want to join and rely on such a trustless environment. To solve these issues, we propose a mechanism in which the distributed model training is performed for detecting malicious nodes. The distributed models are trained and then a unified model is generated at the centralized server. This will not only enhance the accuracy of the unified federated learning model but also preserve the privacy of each cluster because no actual data is sent to fog layer for central model training. We simulate the whole IoT network and for evaluating the performance of our proposed model. The simulation results show that an accuracy of 79% is achieved by our model, indicating that the malicious node is efficiently detected. Furthermore, the precision of our model is 1, which indicates that our model can easily discriminate between the true and false classes.


  • Open Access

    Copy-Move Forgery Detection Technique Using Graph Convolutional Networks Feature Extraction

    by Varun Shinde, Vineet Dhanawat, Ahmad Almogren, Anjanava Biswas, Muhammad Bilal, Rizwan Ali Naqvi, Ateeq Ur Rehman

    In the development of image forensics, detection of Copy-Move Forgery (CMF) has become a major challenge due to the proliferation of image forgery techniques. The CMF is widely utilized to alter the content of the original image to spread false information or to use such forged digital images for illegal purposes e.g. false evidence in the court of law, or to blackmailing any individual. This paper presents a new method for CMF Detection (CMFD) that uses the power of Graph Convolution Networks (GCNs) and its multiple layers with ReLU activation, for CMFD and analysis. The aim to use GCN is due to its ability to improve the feature extraction process by utilizing the spatial and structural affiliation between elements in the digital images. Also, the GCN aims to store information about images and use it to graphically describe images with pixels or image areas as features, spatial and correlation relationships as edges. By pulling data from this image, GCN is able to obtain content rich features that are very powerful at detecting CMF regions. In proposed methodology, we utilized Support Vector machine (SVM) for classification and the binary cross-entropy loss, and the Adam optimizer for improving accuracy. Our scheme successfully achieves high accuracy and is effective in CMFD. We use the MICC F220, and CoMoFoD datasets to test the GCN in our proposed CMFD method. Through much testing and evaluation, we have found that GCN has the tremendous ability for CMFD in digital images in term of accuracy.