Faculty Publications

Now showing 1 - 4 of 4

Role of intensity of emotions for effective personalized video recommendation: A reinforcement learning approach
(Springer Verlag service@springer.de, 2018) Tripathi, A.; Manasa, D.G.; Rakshitha, K.; Ashwin, T.S.; Reddy, G.
Development of artificially intelligent agents in video recommendation systems over past decade has been an active research area. In this paper, we have presented a novel hybrid approach (combining collaborative as well as content-based filtering) to create an agent which targets the intensity of emotional content present in a video for recommendation. Since cognitive preferences of a user in real world are always in a dynamic state, tracking user behavior in real time as well as the general cognitive preferences of the users toward different emotions is a key parameter for recommendation. The proposed system monitors the user interactions with the recommended video from its user interface and web camera to learn the criterion of decision-making in real time through reinforcement learning. To evaluate the proposed system, we have created our own UI, collected videos from YouTube, and applied Q-learning to train our system to effectively adapt user preferences. © Springer Nature Singapore Pte Ltd. 2018
Q-feedan effective solution for the free-riding problem in unstructured P2P networks
(2010) Thampi, S.M.; Sekaran K, C.
This paper presents a solution for reducing the ill effects of free-riders in decentralised unstructured P2P networks. An autonomous replication scheme is proposed to improve the availability and enhance system performance. Q-learning is widely employed in different situations to improve the accuracy in decision making by each peer. Based on the performance of neighbours of a peer, every neighbour is awarded different levels of ranks. At the same time a low-performing node is allowed to improve its rank in different ways. Simulation results show that Q-learning-based free riding control mechanism effectively limits the services received by free-riders and also encourages the low-performing neighbours to improve their position. The popular files are autonomously replicated to nodes possessing required parameters. Due to this improvement of quantity of popular files, free riders are given opportunity to lift their position for active participation in the network for sharing files. Q-feed effectively manages queries from free riders and reduces network traffic significantly. © 2010 S. M. Thampi and C. Sekaran K.
EmoWare: A context-aware framework for personalized video recommendation using affective video sequences
(Institute of Electrical and Electronics Engineers Inc., 2019) Tripathi, A.; Ashwin, T.S.; Guddeti, R.M.R.
With the exponential growth in areas of machine intelligence, the world has witnessed promising solutions to the personalized content recommendation. The ability of interactive learning agents to make optimal decisions in dynamic environments has been proven and very well conceptualized by reinforcement learning (RL). The learning characteristics of deep-bidirectional recurrent neural networks (DBRNN) in both positive and negative time directions has shown exceptional performance as generative models to generate sequential data in supervised learning tasks. In this paper, we harness the potential of the said two techniques and propose EmoWare (emotion-aware), a personalized, emotionally intelligent video recommendation engine, employing a novel context-aware collaborative filtering approach, where the intensity of users' spontaneous non-verbal emotional response toward the recommended video is captured through interactions and facial expressions analysis for decision-making and video corpus evolution with real-time feedback streams. To account for users' multidimensional nature in the formulation of optimal policies, RL-scenarios are enrolled using on-policy (SARSA) and off-policy (Q-learning) temporal-difference learning techniques, which are used to train DBRNN to learn contextual patterns and to generate new video sequences for the recommendation. System evaluation for a month with real users shows that the EmoWare outperforms the state-of-the-art methods and models users' emotional preferences very well with stable convergence. © 2013 IEEE.
Automatic generation control of is-landed micro-grid using integral reinforcement learning-based adaptive optimal control strategy
(Springer Science and Business Media Deutschland GmbH, 2025) Muduli, R.; Jena, D.; Moger, T.
Abstract: Microgrids serve an essential role in the smart grid infrastructure, facilitating the seamless integration of distributed energy resources and supporting the increased adoption of renewable energy sources to satisfy the growing demand for sustainable energy solutions. This paper presents an application of integral reinforcement learning (IRL) algorithm-based adaptive optimal control strategy for automatic generation control of an is-landed micro-grid. This algorithm is a model-free actor-critic method that learns the critic parameters using the recursive least square method. The actor is straightforward and evaluates the action from the critic directly. The robustness of the proposed control technique is investigated under various uncertainties arising from parameter uncertainty, electric vehicle (EV) aggregator, and renewable energy sources. This study incorporates case studies and comparative analyses to demonstrate the control performance of the proposed control strategy. The effectiveness of the technique is evaluated by comparing it with deep Q-learning (DQN) control techniques and PI controllers. The proposed controller significantly improves performance metrics compared to the DQN and PI controllers. It reduces the peak frequency deviation by 6% and 14%, respectively, compared to the DQN and PI controllers. When subjected to multiple-step load disturbances, the proposed controller reduces the mean square error by 28% and 42%, respectively, while lowering both the integral absolute error and the integral time absolute error by 21% and 35% compared to the DQN and PI controllers. Additionally, when operating with renewable energy sources, the proposed controller decreases the standard deviation in the frequency deviation by 17% compared to the DQN controller and 23% compared to the PI controller. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.

Faculty Publications

Browse

Filters

Settings

Sort By

Results per page

Search Results