CM-AMO Seminar | Double Feature
Shuaifeng Li (U-M Physics) and Shriya Sinha/Zecheng You (U-M Physics)
Shuaifeng Li
U-M Physics
Training All-Mechanical Neural Networks for Task Learning Through in Situ Backpropagation
Recent advances unveiled physical neural networks as promising machine learning platforms, offering faster and more energy-efficient information processing. Compared with extensively-studied optical neural networks, the development of mechanical neural networks remains nascent and faces significant challenges, including heavy computational demands and learning with approximate gradients. Here, we introduce the mechanical analogue of in situ backpropagation to enable highly efficient training of mechanical neural networks. We theoretically prove that the exact gradient can be obtained locally, enabling learning through the immediate vicinity, and we experimentally demonstrate this backpropagation to obtain gradient with high precision. With the gradient information, we showcase the successful training of networks in simulations for behavior learning and machine learning tasks, achieving high accuracy in experiments of regression and classification. Our findings, which integrate the theory for training mechanical neural networks and experimental and numerical validations, pave the way for mechanical machine learning hardware and autonomous self-learning material systems.
Shriya Sinha/Zecheng You
U-M Physics
Emerging Conduction Pathways in Semiconducting Bismuth-Antimony Alloys
Bi_{1-x}Sb_x alloys with ~0.07 < x < 0.22 have long been reported as narrow bandgap semiconductors with inverted bands. In addition to hosting topologically protected surface states, the material has certain topological indices allowing it to host conduction pathways via extended defects such as dislocations. At temperatures above 150 K, transport is dominated by thermally activated electrons and holes. The carrier density and their mobilities are determined by performing a two-channel analysis of the longitudinal and transverse magneto-conductivity. A 40 meV band gap is extracted by analyzing the temperature dependence of the carrier densities. ARPES measurements were conducted to explore the bulk bandgap focused on the L-point. We extract electron mobility from a sharp peak of magneto-conductance. From room temperature to cryogenic temperatures, electron mobility increases by more than two orders of magnitude and approaches a remarkably high value of 750,000 cm^2/Vs below 10 K. Because of the high mobility of electrons, we were able to suppress the contribution of the bulk electron by 5-6 orders of magnitude by applying a large magnetic field and discover this presence of a new magnetic-field-independent conducting path of unknown origin.
U-M Physics
Training All-Mechanical Neural Networks for Task Learning Through in Situ Backpropagation
Recent advances unveiled physical neural networks as promising machine learning platforms, offering faster and more energy-efficient information processing. Compared with extensively-studied optical neural networks, the development of mechanical neural networks remains nascent and faces significant challenges, including heavy computational demands and learning with approximate gradients. Here, we introduce the mechanical analogue of in situ backpropagation to enable highly efficient training of mechanical neural networks. We theoretically prove that the exact gradient can be obtained locally, enabling learning through the immediate vicinity, and we experimentally demonstrate this backpropagation to obtain gradient with high precision. With the gradient information, we showcase the successful training of networks in simulations for behavior learning and machine learning tasks, achieving high accuracy in experiments of regression and classification. Our findings, which integrate the theory for training mechanical neural networks and experimental and numerical validations, pave the way for mechanical machine learning hardware and autonomous self-learning material systems.
Shriya Sinha/Zecheng You
U-M Physics
Emerging Conduction Pathways in Semiconducting Bismuth-Antimony Alloys
Bi_{1-x}Sb_x alloys with ~0.07 < x < 0.22 have long been reported as narrow bandgap semiconductors with inverted bands. In addition to hosting topologically protected surface states, the material has certain topological indices allowing it to host conduction pathways via extended defects such as dislocations. At temperatures above 150 K, transport is dominated by thermally activated electrons and holes. The carrier density and their mobilities are determined by performing a two-channel analysis of the longitudinal and transverse magneto-conductivity. A 40 meV band gap is extracted by analyzing the temperature dependence of the carrier densities. ARPES measurements were conducted to explore the bulk bandgap focused on the L-point. We extract electron mobility from a sharp peak of magneto-conductance. From room temperature to cryogenic temperatures, electron mobility increases by more than two orders of magnitude and approaches a remarkably high value of 750,000 cm^2/Vs below 10 K. Because of the high mobility of electrons, we were able to suppress the contribution of the bulk electron by 5-6 orders of magnitude by applying a large magnetic field and discover this presence of a new magnetic-field-independent conducting path of unknown origin.
Building: | West Hall |
---|---|
Event Type: | Workshop / Seminar |
Tags: | Physics, Science |
Source: | Happening @ Michigan from CM-AMO Seminars, Department of Physics |
Events
Featured
Apr
12
Saturday Morning Physics | VanLoo Family SMP: Graduate Student Research Presentations
Otávio Alves and Emilie LaVoie-Ingrahm (U-M Physics)
10:30 AM
170 & 182 Auditoriums
Weiser Hall
Upcoming
Apr
01
CM-AMO Seminar | Entanglement complexity and precision measurement with cold atom qubits
Grant Biedermann (University of Oklahoma)
4:00 PM
340
West Hall
Apr
02
Brown Bag Seminar | Geometric Entropies and their Hamiltonian Flow
Pratik Rath (UC Berkeley)
12:00 PM
3481
Randall Laboratory
Apr
02
Department Colloquium | An introduction to climate engineering
David Keith (University of Chicago)
3:00 PM
340
West Hall