Integrated Calendar

Until recently, pulse EPR was based solely on rectangular pulses. This changed with the introduction
of fast arbitrary waveform generators (AWG) that allow for pulse shaping and phase/frequency modulation at microwave frequencies. Early applications of this technology focused mainly on chirp pulses for broadband excitation and inversion within existing pulse sequences.
In this talk, I will focus on Dynamic Nuclear Polarization with modulated pulse sequences in static solids.
The theoretical description shows remarkable similarities with dipolar recoupling sequences in magic angle spinning (MAS) NMR. In dipolar recoupling, the pulse sequence interferes with the time-dependence of interactions due to the sample spinning. A similar phenomenon takes place in pulsed DNP, where the pulses interfere with the rotation in spin space due to the nuclear Zeeman interaction.
After introducing the theoretical background, I will show results at 0.35 T/15 MHz/9.5 GHz. I will then discuss the implications for pulsed DNP at higher magnetic fields. Finally, I show and propose experiments to make use of DNP within the context of pulse EPR, i.e. for detecting hyperfine coupled nuclei in the vicinity of unpaired electrons

One of the puzzling phenomena in deep learning, is that neural networks tend to generalize well even when they are highly overparameterized. Recent works linked this behavior with implicit biases of the algorithms used to train networks (like SGD). Here we analyze the implicit bias of SGD from the standpoint of minima stability, focusing on shallow ReLU networks trained with a quadratic loss. Specifically, it is known that SGD can stably converge only to minima that are flat enough w.r.t. its step size. Here we show that this property enforces the predictor function to become smoother as the step size increases, thus significantly regularizing the solution. Furthermore, we analyze the representation power of stable solutions. Particularly, we prove a depth-separation result: There exist functions that cannot be approximated by depth-2 networks corresponding to stable minima, no matter how small the step size is taken to be, but which can be implemented with depth-3 networks corresponding to stable minima. We show how our theoretical findings explain behaviors observed in practical settings.
(Joint works with Rotem Mulayoff, Mor Shpigel Nacson, Greg Ongie, Daniel Soudry).