Assessing Perceived Fairness in Machine Learning (ML) Process: A Conceptual Framework 

In ML applications, “unfairness” can be caused by bias in the data, curation process, erroneous assumptions, and implicit bias rendered within the algorithmic development process. As ML applications come into broader use, developing fair ML applications is critical. Assessing fairness and developing fair ML applications has become important in the era of Responsible AI in practice in research, industry, and academia. However, a literature survey suggests that fairness in ML is very subjective, and there is no coherent way to describe the fairness of AI/ML processes and applications. To better understand the perception of fairness in the ML process, we conducted virtual focus groups with developers, reviewed prior literature, and integrated notions of justice theory to propose that perceived fairness is a multidimensional concept. In this paper, we will explore the initial outcomes of this effort.

Click on the left image for enlarged view**

Mishra, A., Khazanchi, D. (2023). Assessing Perceived Fairness in Machine Learning (ML) Process: A Conceptual Framework, NeurIPS workshop on Algorithmic Fairness through the Lens of Time, New Orleans

A Proposal for Research on the Application of AI/ML in ITPM: Intelligent Project Management  

According to the market research firm Tractica, the global artificial intelligence software market is forecast to grow to 126 billion by 2025. Additionally, the Gartner group predicts that during the same time as much as 80% of the routine work ,  which represents the bulk of human hours spent in today's project management (PM) activities, can be eliminated because of collaboration between humans and smart machines. Today's PM practices rely heavily on human input. However, that is not the optimum use of the human project manager's intuitive, innovative, and creative abilities. Many aspects of a project manager's work could be managed by machines that utilize AI/ML approaches to address nonroutine and predictive tasks. This paper describes IT project management (ITPM) processes and associated tasks and identifies the AI/ML approaches that can support them.

Mishra, A., Khazanchi, D., Tripathi, A. (2022). A Proposal for Research on the Application of AI/ML in ITPM: Intelligent Project Management, International Journal of Information Technology Project Management (IJITPM) 

Exploring Bitslicing Architectures for Enabling FHE-assisted Machine Learning 

Homomorphic Encryption (HE) is the ultimate tool for performing secure computations, even in untrusted environments. Application of HE for Deep Learning (DL) inference is an active area of research, given the fact that DL models are often deployed in untrusted environments (e.g. third-party servers) yet inferring on private data. However, existing HE libraries (somewhat (SWHE), leveled (LHE) or fully homomorphic (FHE)) suffer from extensive computational and memory overhead. Few performance optimized high-speed homomorphic libraries are either suffering from certain approximation issues leading to decryption errors or proven to be insecure according to recent published attacks. In this paper, we propose architectural tricks to achieve performance speedup for encrypted DL inference developed with exact HE schemes without any approximation or decryption error in homomorphic computations. The main idea is to apply quantization and suitable data packing in the form of bitslicing to reduce the costly noise handling operation, Bootstrapping while achieving a functionally correct and highly parallel DL pipeline with a moderate memory footprint. Experi-mental evaluation on the MNIST dataset shows a significant (37X) speedup over the non-bitsliced versions of the same architecture. Low memory bandwidths (700MB) of our design pipelines further highlight their promise towards scaling over larger gamut of Edge-AI analytics use cases.

S. Sinha et al., "Exploring Bitslicing Architectures for Enabling FHE-assisted Machine Learning," in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2022, doi: 10.1109/TCAD.2022.3204909. URL: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9920696&isnumber=6917053

Soumik Sinha, Sayandeep Saha, Manaar Alam, Ayantika Chatterjee, Anoop Mishra, Deepak Khazanchi and Debdeep Mukhopadhyay, Exploring Bitslicing Architectures for Enabling FHE-assisted Machine Learning, Accepted at CASES: International Conference on Compilers, Architectures, and Synthesis for Embedded Systems, Embedded Systems Week, OCTOBER 07-14, 2022, HYBRID-SHANGHAI; Youtube Link: https://youtu.be/VHsj-XYniHw

Glaucoma patients experience difficulty with a number of daily activities, including reading, mobility, and driving. Visual search is instrumental to rapidly localizing target objects among competing objects within our environment. While numerous studies have evaluated factors that influence visual search performance, limited evidence is available for how these factors affect visual search in glaucoma patients. Our core hypothesis is that environmental scene context and image characteristics fundamentally influence visual search performance in glaucoma. We propose a conceptual model for evaluating visual search performance in glaucoma patients using artificial intelligence and image processing techniques. Here, the Berlin object in the scene database (BOiS) is used. This database contains 130 high-resolution naturalistic images intended to evaluate visual search performance in a real-world setting. We proposed a model that will improve our understanding of glaucoma-related impairments and their impact on visual search will help to improve the quality of life for a significant portion of the population.

Paper Link: Mishra, A., Anderson, D., Belcher, S., & Khazanchi, D. (2020). Evaluating visual search in glaucoma using deep learning. In 26th Americas Conference on Information Systems, AMCIS 2020. Association for Information Systems. 

AIforSocialGood Symposium at CMU Pittsburgh: https://youtu.be/f9EDTUULtR4

Research presentation: https://digitalcommons.unomaha.edu/srcaf/2020/schedule/179/

Intelligent Human-Aware Decision-Making for Semi-Autonomous Human Rehabilitation Assistance Using Modular Robots

Modular Self-reconfigurable Robots (MSRs) are robots that can adapt their shape and mobility while performing their operations. In this research, we consider the locomotion or gait adaptation problem for MSRs in the context of an MSR called MARIO (Modular Robots for Assistance in Robust and Intelligent Operations). The objective of MARIO is to assist patients with spinal cord injury in performing daily living tasks such as fetching out-of-reach objects, e.g., TV remotes, pillows or laundry baskets which leads towards rehabilitation assistance. To address the locomotion problem of MARIO, we propose to use a machine learning-based framework called deep reinforcement learning (DRL) that enables MARIO to autonomously adapt its actions while performing navigation to move towards and reach the object being fetched and bring it back to the human user. We extend the DRL framework with a semi-autonomous decision-making technique called shared autonomy where MARIO intelligently decides to relinquish control to the human when the human wants to intervene to maneuver MARIO during its movements to fetch the object. Our techniques have been validated within an accurately simulated version of MARIO on the Webots robot simulator within daily living environments such as a kitchen and a living room. Our results show that as compared to a reactive controller for MARIO’s motion control, using deep reinforcement learning improves the average time required by MARIO to perform fetching tasks by up to 70 percent when MARIO has to travel longer distances, around 11 meters, to fetch an object, and by up to 85 percent for shorter fetching distances, around 6 meters. Corresponding improvements in average time while integrating shared autonomy with deep reinforcement learning are up to 18 percent when MARIO has to travel longer distances around 11 meters, and by up to 30 percent for shorter fetching distances around 6 meters.

This work was successfully defended in May, 2019 as a Master's Thesis at University of Nebraska at Omaha; Graduate Thesis Advisor: Dr. Prithviraj Dasgupta.; Project Website:  Link

Dasgupta, P., Mishra, A., Nelson, C., Burnfield, J. (2019) “Towards Intelligent Semi-Autonomous Control of a Modular Robot for Human Mobility Assistance,” 2019 Do Good Robotics Symposium (DGRS'19), October 3-4, 2019, College Park, MD. 

Mishra, A. (2019). Intelligent Human-aware Decision Making for Semi-autonomous Human Rehabilitation Assistance Using Modular Robots (Link: Master’s dissertation, University of Nebraska at Omaha).

Mishra, A, Dasgupta, P., Intelligent & Human-Aware Decision Making for Semi-Autonomous Human Rehabilitation Assistance using Modular Robots, Nebraska Academy of sciences & NASA Nebraska space grant 2018(Poster) [Conference program link]

Mishra, A, Dasgupta, P., Intelligent & Human-Aware Decision Making for Semi-Autonomous Human Rehabilitation Assistance using Modular Robots, Student Research and Creativity Fair, Office of Creative and Research Activity, University of Nebraska at Omaha, 2018(Presentation) (Link)

Mishra, A, "Intelligent and Human-Aware Decision Making for Semi-Autonomous Human Rehabilitation Assistance using Modular Robots", CS Workshop 2018, The University of Nebraska at Omaha  (Link)