I am happy to announce that I have been awarded a new EPSRC grant – technically it is UKRI & RCN (Research Council of Norway) project – UKRI-RCN: Exploiting the dynamics of self-timed machine learning hardware (ESTEEM).
I am very excited to work on it with my two Newcastle colleagues Rishad Shafik and Domenico Balsamo, Uni of Agder colleague Ole-Christoffer Granmo, and in close collaboration with PragmatIC, Mignon and CFT.
More details on this project can be found here.
I had a pleasure to present a keynote talk at the 13th International Conference Dependable Systems, Services and Technologies (DESSERT 2023), held in Greece, Athens, October 13-15, 2023, in a hybrid mode.
The talk’s topic was “Tsetlin Machines: stepping towards energy-efficient, explainable and dependable AI” https://www.dessert-conf.org/dessert-2023/alex-yakovlev/
The PDF of the slides can be found here
Victor Ilyich Varshavsky was born today, on 23rd February, 90 years ago. Victor Varshavsky is a pioneer of automata theory, aperiodic (aka self-timed) circuits and systems https://en.wikipedia.org/wiki/Asynchronous_circuit, and collective behaviour of automata. In the 1960s, being a close colleague and friend of Mikhail Tsetlin, Victor laid foundation to the theory of learning automata and machine intelligence, which find their way today to modern methods of machine learning – such as Tsetlin Machine: https://en.wikipedia.org/wiki/Tsetlin_machine
You can read about Victor Varshavsky’s contributrions in this document:https://web.cecs.pdx.edu/~mperkows/CLASS_573/Asynchr_Febr_2007/M.pdf
I am immensely proud to be one of his disciples.
Last week I gave a public lecture “”Data-driven computing (or Liberating computing from memory walls)”, at Technical University of Vienna, Austria, where I am acting as Guest Professor for 2022.
The lecture was on my relatively recent ideas of bringing machine learning into computing at different scales and levels of abstraction, basically making it a commodity that can be introduced for improving the quality of computing from many aspects, in particular performance and use of energy.
The advert of the lecture can be found here: https://informatics.tuwien.ac.at/news/2199
There is also a recording of the lecture available here: https://tube1.it.tuwien.ac.at/w/ebJrRwrJP2ozpsoWAyfy3T
I predict that during and following this period of COVID pandemics, we will witness a significant rise in of interest and some kind of renaissance of mathematics and other STEM subjects. You might ask, why?
Well, let’s look back into history. The development of many mathematical ideas and forms such as mathematical series like geometric series, Fibonacci series, theory of probability etc. were the result of people observing various processes in time or frequency domains during those epidemics like plague, cholera and so on, that took place in the past centuries.
Now, you can see how many smart people are doing home schooling and teach their kids to look at the geometric series and exponential and power laws of the proliferation of virus. A 7-8 year ol kid can have a good grasp of the series based models because he or she could witness its manifestation (sadly, but) in vivo.
So, being an academic in Engineering and curious in anything natural, I hope there will be more students doing Maths, Sciences and Engineering after that ….
I was invited to University of Agder, in the South of Norway (in a nice town called Grimstad, famous for the presence of Henrik Ibsen and Knut Hamsun), to present my vision on what kind of hardware do we need for pervasive AI. This presentation was part of a workshop organised by Prof Ole-Christoffer Granmo, Director of CAIR, on the occasion of the grant opening of CAIR – https://cair.uia.no
In my presentation I emphasized the following points:
- Pervasive Intelligence requires reconsidering many balances:
– Between software and hardware
– Between power and compute
– Between analogand digital
– Between design and fabrication and maintenance
- Granulation phenomenon: Granularity of power, time, data and function
- Main research questions:
– Can we granulate intelligence to minimum?
– What is the smallest level at which we can make cyber-systems learn in terms of power, time, data and function?
- Grand challenge for pervasive hardware AI:
To enable electronic components with an ability to learn and compute in real-life environments with real-power and in real-time
- Research Hypothesis:
We should design systems that are energy-modulated and self-timed, with maximally distributed learning capabilities
I put a strong hypothesis on the role of using Tsetlin Automata (Automata with Linear Tactics) for building electronics with high-granularity learning capabilities.
The key elements of the proposed approach are:
- Event-driven, robust to power and timing fluctuations
- Decentralised TsetlinAutomata (TAs) for learning on demand
- Mixed digital-analogcompute where elements are enabled and controlled by individual TAs
- Natural approximation in its nature, both in learning and compute
- Asynchronous logic for h/w implementation
The full set of my slides is here: https://www.staff.ncl.ac.uk/alex.yakovlev/home.formal/talks/AlexYakovlev-AI%20Hardware-070219.version3.pdf