Artificial intelligence for the Sciences (AI4theSciences) is an innovative, interdisciplinary and intersectoral PhD programme, led by Université Paris Sciences et Lettres (PSL) and co-funded by the European Commission.
Supported by the European innovation and research programme Horizon 2020-Marie Sklodowska-Curie Actions, AI4theSciences is uniquely shaped to train a new generation of researchers at the highest academic level in their main discipline (Physics, Engineering, Biology, Human and Social Sciences) and master the latest technologies in Artificial Intelligence and Machine Learning which apply in their own field.
26 doctoral students will join the PSL university's doctoral schools in 2 academic cohorts to carry out work on subjects suggested and defined by PSL's scientific community.
The 2020 call will offer up to 15 PhD positions on 24 PhD research projects. The candidates will be recruited through HR processes of high standard, based on transparency, equal opportunities and excellence.
Description of the PhD subject : Towards neuromorphic computing on quantum many-body architectures
Context - Motivation
Artificial intelligence (AI) algorithms today are coded into silicon based computer architectures. However, the promise of AI cannot reach its full potential until the underlying architectures more closely resemble the synapses and neurons of the brain.
While many materials are considered candidates for mimicking synapses, only a handful of materials, all quantum many-body systems, are considered viable for mimicking neurons.
In these materials, electrons clump into richly textured shapes. These multiscale structures likely hold the key to neuron-like properties as the materials become highly susceptible to stimuli like electrical spikes in brains.
To understand these multiscale structures we propose combining for the first time three high spatial and temporal resolution surface probes optical microscope, scanning near-field optical microscopy (sNOM), and scanning tunneling microscopy (STM) and robust image analysis based on new machine learning (ML) and statistical mechanics theoretical methods.
This full experimental / theoretical approach will help us to characterize and learn to control the properties of these neuromorphic quantum materials in order to facilitate their eventual incorporation into neuromorphic computer architectures while advancing the field of quantum many-body systems.
This is an international collaboration between the supervisors’s experimental research group and the co-supervisor’s theoretical research group.
The theoretical techniques including image processing, cluster analysis techniques, machine learning techniques, and resistor network methods.
They will be applied to interpret detailed spatially resolved experimental data from optical microscope movies, sNOM, and possibly STM applied to the candidate neuristor materials VO2, V2O3 and NdNiO3.
Over the last three years, the ESPCI - PSL in Paris has set up a nanoscale platform unique in France to characterize quantum materials.
This project will use two state of the art probes in this platform : a low temperature STM and a Linkam self-focusing optical microscope.
The first allows surface sample mapping at atomic resolution on metal and insulating materials using an atomic force microcope (AFM) position tip mode.
The second allows micron resolution image movies at 100ms intervals to probe the rich temporal dynamics in these neuromorphic quantum materials.
These will be complemented by sNOM measurements with 20nm resolution, through an ongoing ICAM / I2CAM collaboration with Dmitri Basovs’s group at Columbia University.
By overlapping the datasets of all three probes, we will map out for the first time six decades of spatial range to 100microns) in these neuromorphic materials.
These multirange maps will permit not only (i) static fine structure analysis in the image analysis process but also the identification of (ii) dynamic fine changes when these quantum materials are placed in neuromorphic circuits which use electrical spike signal propagation.
Scientific Objectives, Methodology & Expected results
We use computational techniques to predict expected
Many of the materials form intricate structures with fractal boundaries and interiors, and the cluster analysis techniques quantify these and other geometric measures to compare with theoretical models.
The ML algorithms do not require this intermediate step of interpretation. Rather, we use simulations to predict thousands of potential patterns from each type of model and, under supervised learning conditions, we use those patterns to train a convolutional neural network to recognize which model generated which image.
Once trained, the neural network will be applied to images derived from experimental data. The combination of these two complementary theoretical techniques will allow us to develop detailed models of the fundamental physics driving the intricate pattern formation in these quantum materials.
A combination of cluster analysis and ML analysis mentioned above with a resistor network analysis will be used to identify any fine changes in the fractal pattern formation.
This will allow us for the first time to visualize and follow the creation of percolation paths, giving us the opportunity to control the properties of these neuromorphic quantum materials.
The ultimate goal of our research is to realize the potential of quantum materials to build disruptive computer architectures capable of mimicking the brain, including synapses and neurons .
While many materials are considered candidates for mimicking synapses, only a handful of materials are considered viable for mimicking neurons.
In this project, we will combine optical experimental techniques with AI and informatics theoretical methods to characterize and learn to control the properties of four of these materials, VO2, V2O3, NdNiO3, and LCMO, in order to facilitate their eventual incorporation into neuromorphic computers.
The PhD candidate will be working on experimental setups and data treatment using machine learning. A trip to Purdue University (IN, USA) in the team of Prof.
Erica W. Carlson will be planned during this project. The travel schedule and length will be determined according to experimental work progress and data available.
Lionel Aigouy and Alexandre Zimmers
Created in 2012, Université PSL is aiming at developing interdisciplinary training programmes and science projects of excellence within its members.
Its 140 laboratories and 2,900 researchers carry out high-level disciplinary research, both fundamental and applied, fostering a strong interdisciplinary approach.
The scope of Université PSL covers all areas of knowledge and creation (Sciences, Humanities and Social Science, Engineering, the Arts).
Its eleven component schools gather 17,000 students and have won more than 200 ERC. PSL has been ranked 36th in the 2020 Shanghai ranking (ARWU).
Required Research Experiences
Skills / Qualifications