Luca D

Building credibility using machine learning algorithms

 Luca Discepolo, postgraduate student at the University of Bristol

Credibility bursary supported attendance at: Computational Neuroscience, Neurotechnology and Neuro-inspired AI Autumn School, University of Ulster, 2023

What are some of the challenges with credibility that particularly affect your work?


One of the significant challenges in my work involves methodologies and tools utilized in my research. In my PhD project, I focused on investigating synapse development within associative and sensory areas. Specifically, my focus was on characterizing the density and features of synaptic protein aggregates, known as puncta, employing a plugin designed to detect these puncta based on predefined thresholds and diameters.


To ensure the accuracy of the detection tool, I undertook a validation process by comparing its performance against manual detection methods. While manual detection proved effective and reliable in validating the tool's functionality, I recognize the importance of further bolstering the credibility of my findings by comparing the results obtained with alternative automated systems. For instance, integrating the outputs of a machine learning-based image analysis tool could serve to augment the robustness and validity of my research outcomes. 


By employing multiple verification methods and cross-referencing with different automated systems, I aim to enhance the trustworthiness and credibility of my research findings in the field of synapse development.


What are some of the key learnings you’ve taken away from the Autumn School that will strengthen credibility?


Attending the Autumn School in  Computational Neuroscience and Artificial Intelligence in Derry, thanks to the bursary granted by the BNA, has been immensely beneficial in enhancing my understanding of various machine learning algorithms that can be applied in the analysis of large datasets, but also the methods that are used to validate them, for instance cross-validation and bootstrapping.


The technical lectures were nicely complemented with interesting seminars held by experts in the field of ethics and regulatory issues in neuroscience. These seminars delved into critical topics, such as data privacy and ownership. I gained insights about the unique challenges that AI inheritably yields through the discussion with my peers.


How are you planning to apply/share this knowledge in your current work?


In addressing the challenge of comparing results across different projects or brain areas without a common framework, I aim to establish a standardized pipeline in my current work. This involves developing a consistent approach for data processing, analysis, and interpretation, ensuring comparability across studies. By implementing a unified methodology, I can facilitate seamless integration of findings and meaningful comparisons with other research endeavors.


Additionally, I plan to disseminate this knowledge through workshops, presentations, and collaborations within the scientific community, promoting methodological transparency and reproducibility.


Through these efforts, I seek to enhance the reliability and credibility of my research outcomes while contributing to the broader advancement of neuroscience methodologies.



Share by: