A series of III-V ternary and quarternary digital alloy avalanche photodiodes (APDs) have recently been seen to exhibit very low excess noise. Using band inversion of an environment-dependent atomistic tight binding description of short period superl
attices, we argue that a combination of increased effective mass, minigaps and band split-off are primarily responsible for the observed superior performance. These properties significantly limit the ionization rate of one carrier type, either holes or electrons, making the avalanche multiplication process unipolar in nature. The unipolar behavior in turn reduces the stochasticity of the multiplication gain. The effects of band folding on carrier transport are studied using the Non-Equilibrium Greens Function Method that accounts for quantum tunneling, and Boltzmann Transport Equation model for scattering. It is shown here that carrier transport by intraband tunneling and optical phonon scattering are reduced in materials with low excess noise. Based on our calculations, we propose five simple inequalities that can be used to approximately evaluate the suitability of digital alloys for designing low noise photodetectors. We evaluate the performance of multiple digital alloys using these criteria and demonstrate their validity.
To quantitatively and intuitively explore the generalization ability of pre-trained language models (PLMs), we have designed several tasks of arithmetic and logical reasoning. We both analyse how well PLMs generalize when the test data is in the same
distribution as the train data and when it is different, for the latter analysis, we have also designed a cross-distribution test set other than the in-distribution test set. We conduct experiments on one of the most advanced and publicly released generative PLM - BART. Our research finds that the PLMs can easily generalize when the distribution is the same, however, it is still difficult for them to generalize out of the distribution.
Imitation learning aims to extract knowledge from human experts demonstrations or artificially created agents in order to replicate their behaviors. Its success has been demonstrated in areas such as video games, autonomous driving, robotic simulatio
ns and object manipulation. However, this replicating process could be problematic, such as the performance is highly dependent on the demonstration quality, and most trained agents are limited to perform well in task-specific environments. In this survey, we provide a systematic review on imitation learning. We first introduce the background knowledge from development history and preliminaries, followed by presenting different taxonomies within Imitation Learning and key milestones of the field. We then detail challenges in learning strategies and present research opportunities with learning policy from suboptimal demonstration, voice instructions and other associated optimization schemes.
Federated learning (FL) is an emerging paradigm for machine learning, in which data owners can collaboratively train a model by sharing gradients instead of their raw data. Two fundamental research problems in FL are incentive mechanism and privacy p
rotection. The former focuses on how to incentivize data owners to participate in FL. The latter studies how to protect data owners privacy while maintaining high utility of trained models. However, incentive mechanism and privacy protection in FL have been studied separately and no work solves both problems at the same time. In this work, we address the two problems simultaneously by an FL-Market that incentivizes data owners participation by providing appropriate payments and privacy protection. FL-Market enables data owners to obtain compensation according to their privacy loss quantified by local differential privacy (LDP). Our insight is that, by meeting data owners personalized privacy preferences and providing appropriate payments, we can (1) incentivize privacy risk-tolerant data owners to set larger privacy parameters (i.e., gradients with less noise) and (2) provide preferred privacy protection for privacy risk-averse data owners. To achieve this, we design a personalized LDP-based FL framework with a deep learning-empowered auction mechanism for incentivizing trading gradients with less noise and optimal aggregation mechanisms for model updates. Our experiments verify the effectiveness of the proposed framework and mechanisms.
This paper introduces the SemEval-2021 shared task 4: Reading Comprehension of Abstract Meaning (ReCAM). This shared task is designed to help evaluate the ability of machines in representing and understanding abstract concepts. Given a passage and th
e corresponding question, a participating system is expected to choose the correct answer from five candidates of abstract concepts in a cloze-style machine reading comprehension setup. Based on two typical definitions of abstractness, i.e., the imperceptibility and nonspecificity, our task provides three subtasks to evaluate the participating models. Specifically, Subtask 1 aims to evaluate how well a system can model concepts that cannot be directly perceived in the physical world. Subtask 2 focuses on models ability in comprehending nonspecific concepts located high in a hypernym hierarchy given the context of a passage. Subtask 3 aims to provide some insights into models generalizability over the two types of abstractness. During the SemEval-2021 official evaluation period, we received 23 submissions to Subtask 1 and 28 to Subtask 2. The participating teams additionally made 29 submissions to Subtask 3. The leaderboard and competition website can be found at https://competitions.codalab.org/competitions/26153. The data and baseline code are available at https://github.com/boyuanzheng010/SemEval2021-Reading-Comprehension-of-Abstract-Meaning.
The unprecedented wide bandgap tunability (~1 eV) of Al$_x$In$_{1-x}$As$_y$Sb$_{1-y}$ latticed-matched to GaSb enables the fabrication of photodetectors over a wide range from near-infrared to mid-infrared. In this paper, the valence band-offsets in
AlxIn1-xAsySb1-y with different Al compositions are analyzed by tight-binding calculations and X-ray photoelectron spectroscopy (XPS) measurements. The observed weak variation in valence band offsets is consistent with the lack of any minigaps in the valence band, compared to the conduction band.
Avalanche photodiodes fabricated from AlInAsSb grown as a digital alloy exhibit low excess noise. In this paper, we investigate the band structure-related mechanisms that influence impact ionization. Band-structures calculated using an empirical tigh
t-binding method and Monte Carlo simulations reveal that the mini-gaps in the conduction band do not inhibit electron impact ionization. Good agreement between the full band Monte Carlo simulations and measured noise characteristics is demonstrated.
The disordered microphases that develop in the high-temperature phase of systems with competing short-range attractive and long-range repulsive (SALR) interactions result in a rich array of distinct morphologies, such as cluster, void cluster and per
colated (gel-like) fluids. These different structural regimes exhibit complex relaxation dynamics with significant relaxation heterogeneity and slowdown. The overall relationship between structure and configurational sampling schemes, however, remains largely uncharted. In this article, the disordered microphases of a schematic SALR model are thoroughly characterized, and structural relaxation functions adapted to each regime are devised. The sampling efficiency of various advanced Monte Carlo (MC) sampling schemes--Virtual-Move (VMMC), Aggregation-Volume-Bias (AVBMC) and Event-Chain (ECMC)--is then assessed. A combination of VMMC and AVBMC is found to be computationally most efficient for cluster fluids and ECMC to become relatively more efficient as density increases. These results offer a complete description of the equilibrium disordered phase of a simple microphase former as well as dynamical benchmarks for other sampling schemes.
A silicon quantum photonic circuit was proposed and demonstrated as an integrated quantum light source for telecom band polarization entangled Bell state generation and dynamical manipulation. Biphoton states were firstly generated in four silicon wa
veguides by spontaneous four wave mixing. They were transformed to polarization entangled Bell states through on-chip quantum interference and quantum superposition, and then coupled to optical fibers. The property of polarization entanglement in generated photon pairs was demonstrated by two-photon interferences under two non-orthogonal polarization bases. The output state could be dynamically switched between two polarization entangled Bell states, which was demonstrated by the experiment of simplified Bell state measurement. The experiment results indicate that its manipulation speed supported a modulation rate of several tens kHz, showing its potential on applications of quantum communication and quantum information processing requiring dynamical quantum entangled Bell state control.
This paper focuses on finding reinforcement learning policies for control systems with hard state and action constraints. Despite its success in many domains, reinforcement learning is challenging to apply to problems with hard constraints, especiall
y if both the state variables and actions are constrained. Previous works seeking to ensure constraint satisfaction, or safety, have focused on adding a projection step to a learned policy. Yet, this approach requires solving an optimization problem at every policy execution step, which can lead to significant computational costs. To tackle this problem, this paper proposes a new approach, termed Vertex Networks (VNs), with guarantees on safety during exploration and on learned control policies by incorporating the safety constraints into the policy network architecture. Leveraging the geometric property that all points within a convex set can be represented as the convex combination of its vertices, the proposed algorithm first learns the convex combination weights and then uses these weights along with the pre-calculated vertices to output an action. The output action is guaranteed to be safe by construction. Numerical examples illustrate that the proposed VN algorithm outperforms vanilla reinforcement learning in a variety of benchmark control tasks.
