Hi there! My name is Dovydas. I’m a researcher based at University College London (UCL), where I investigate cybersecurity threats to machine learning hardware. Outside university, I tutor GCSE, A-level, and IBDP students in mathematics and physics.

Find Me




I work with hardware that aims to make machine learning faster and more power-efficient. Specifically, I focus on circuits made up of memristors, which are devices that act a lot like variable resistors. The ability to change their resistance means that they are good for encoding information (as long as you don’t require perfect precision!). Arranging them in a special way also enables you to perform some mathematical operations widely used in machine learning (e.g. matrix multiplication) without moving those encoded data thus saving time and increasing power efficiency.

My current focus is cybersecurity threats to such machine learning hardware. Even in conventional (transistor-based) computers, adversarial attacks may be used to confuse machine learning systems, e.g. an attacker might trick an autonomous vehicle into thinking that “60 mph” on a street sign is actually “85 mph” instead. This can cause a lot of problems, thus it is important to understand

  • whether memristor-based hardware is as susceptible to such threats
  • whether effective defense strategies exist


PhD in Electronic Engineering (2018–2022, UCL)

I investigated memristor-based hardware accelerators for machine learning. Artificial neural networks and cognitive computing more generally have achieved amazing results over the last decade. However, conventional computers, which separate memory and computing units, are not well suited to the computations that neural networks perform. As a result, a lot of time and energy is wasted. One of the suggested alternatives is an emerging technology—memristor-based hardware. Although vastly more efficient, memristor-based implementations, compared to traditional digital computers, exhibit more variability. This could present problems.

I used experimental data and simulations to

I explain the rationale behind the research in less technical terms in the following posts:


Original research
  • D. Joksas, E. Wang, N. Barmpatsalos, W. H. Ng, A. J. Kenyon, G. A. Constantinides, and A. Mehonic, “Nonideality-aware training for accurate and robust low-power memristive neural networks,” Advanced Science, vol. 9, no. 17, p. 2105784, 2022. doi:10.1002/advs.202105784
  • D. Joksas, P. Freitas, Z. Chai, W. H. Ng, M. Buckwell, C. Li, W. D. Zhang, Q. Xia, A. J. Kenyon, and A. Mehonic, “Committee machines—a universal method to deal with non-idealities in memristor-based neural networks,” Nature Communications, vol. 11, no. 1, 2020. doi:10.1038/s41467-020-18098-0
  • A. Mehonic, D. Joksas, W. H. Ng, M. Buckwell, and A. J. Kenyon, “Simulation of inference accuracy using realistic RRAM devices,” Frontiers in Neuroscience, vol. 13, p. 593, 2019. doi:10.3389/fnins.2019.00593
  • D. Joksas and A. Mehonic, “badcrossbar: A Python tool for computing and plotting currents and voltages in passive crossbar arrays,” SoftwareX, vol. 12, p. 100617, 2020. doi:10.1016/j.softx.2020.100617
  • D. Joksas, A. AlMutairi, O. Lee, M. Cubukcu, A. Lombardo, H. Kurebayashi, A. J. Kenyon, and A. Mehonic, “Memristive, spintronic, and 2D-materials-based devices to improve and complement computing hardware,” Advanced Intelligent Systems, vol. 4, no. 8, p. 2200068, 2022. doi:10.1002/aisy.202200068
BEng Electronic Engineering (2015–2018, UCL)

The modules that I found the most intellectually rewarding were Photonics and Communication Systems (ELEC215P), Control Systems I (ELEC3003), Numerical Methods (ELEC3030), Quantum Physics (PHAS2222), and Atomic and Molecular Physics (PHAS2224).


Original research
  • A. J. Kenyon, M. S. Munde, W. H. Ng, M. Buckwell, D. Joksas, and A. Mehonic, “The interplay between structure and function in redox-based resistance switching,” Faraday Discussions, vol. 213, pp. 151–163, 2019. doi:10.1039/C8FD00118A

Erdős Number


My Favorite…


  • A Clockwork Orange [1971]
  • Adaptation [2002]
  • American Beauty [1999]
  • Back to the Future [1985]
  • Barry Lyndon [1975]
  • Blue Velvet [1986]
  • Boogie Nights [1997]
  • Fargo [1996]
  • Full Metal Jacket [1987]
  • Goodfellas [1990]
  • Interstellar [2014]
  • La piel que habito [2011]
  • Leaving Las Vegas [1995]
  • Lost Highway [1997]
  • Mulholland Drive [2001]
  • Once Upon a Time in the West [1968]
  • Phantom Thread [2017]
  • Reservoir Dogs [1992]
  • Taxi Driver [1976]
  • The Departed [2006]
  • The Florida Project [2017]
  • The Irishman [2019]
  • The King of Comedy [1983]
  • The Master [2012]
  • The Prestige [2006]
  • The Truman Show [1998]
  • There Will Be Blood [2007]
  • Threads [1984]
  • Trainspotting [1996]
  • Wild at Heart [1990]

Songs (performances)

  • 50 Ways to Leave Your Lover [Paul Simon, 1975]
  • Any Way You Want It [Journey, 1980]
  • Ballad of a Thin Man [Bob Dylan, 1965]
  • Eloise [Barry Ryan, 1968]
  • First We Take Manhattan [Leonard Cohen, 1986]
  • Happens to the Heart [Leonard Cohen, 2019 (2016)]
  • Heroes and Villains [The Beach Boys, 1967]
  • Hurricane [Bob Dylan, 1976]
  • Island in the Sun [Harry Belafonte, 1957]
  • Jessie's Girl [Rick Springfield, 1981]
  • Lah-Di-Dah [Jake Thackray, 1991]
  • Mr. Sandman [The Chordettes, 1954]
  • Paint It, Black [The Rolling Stones, 1966]
  • People Are People [Depeche Mode, 1984]
  • Red Light Spells Danger [Billy Ocean, 1976]
  • She's a Rainbow [The Rolling Stones, 1967]
  • Stop! In the Name of Love [The Supremes, 1965]
  • Strawberry Fields Forever [The Beatles, 1967]
  • Summertime [Billy Stewart, 1966]
  • Waiting Around to Die [Townes Van Zandt, 1969]
  • War [Edwin Starr, 1970]
  • You And Me [Spargo, 1980]