Search

Tennis, like other games and sports, is governed by rules, including the rules that determine the winner of points, games, sets, and matches. If the two players are equally skilled -- each has the same probability of winning a point when serving or when receiving -- we show that each has an equal chance of winning games, sets, and matches, whether or not sets go to a tiebreak. However, in a women's match that is decided by 2 out of 3 sets, and a men's match that is decided by 3 out of 5 sets, it is possible that the player who wins the most games may not be the player who wins the match. We calculate the probability that this happens and show that it has actually occurred -- most notably, in the 2019 men's Wimbledon final between Novak Djokovic and Roger Federer, which took almost five hours to complete and is considered one of the greatest tennis matches ever (Djokovic won). We argue that the discrepancy between the game winner and the match winner, when it occurs, should be resolved by a Grand Tiebreak (GT) -- played according to the rules of tiebreaks in sets -- because each player has a valid claim to being called the rightful winner. A GT would have the salutary effect of -- even every point -- lest he/she win in sets but lose more games. This would make competition keener throughout a match and probably decrease the need for a GT, because the game and set winner would more likely coincide when the players fight hard for every point.

The point cloud is a flexible representation for a wide variety of data types, and is a particularly natural fit for the 3D conformations of molecules. Extant molecule embedding/representation schemes typically focus on internal degrees of freedom, ignoring the global 3D orientation. For tasks that depend on knowledge of both molecular conformation and 3D orientation, such as the generation of molecular dimers, clusters, or condensed phases, we require a representation which is provably complete in the types and positions of atomic nuclei and roto-inversion equivariant with respect to the input point cloud. We develop, train, and evaluate a new type of autoencoder, molecular O(3) encoding net (Mo3ENet), for multi-type point clouds, for which we propose a new reconstruction loss, capitalizing on a Gaussian mixture representation of the input and output point clouds. Mo3ENet is end-to-end equivariant, meaning the learned representation can be manipulated on O(3), a practical bonus for downstream learning tasks. An appropriately trained Mo3ENet latent space comprises a universal embedding for scalar and vector molecule property prediction tasks, as well as other downstream tasks incorporating the 3D molecular pose., Comment: 16 pages, 8 figures, including main text, bibliography and supplemental material

Identifying local structural motifs and packing patterns of molecular solids is a challenging task for both simulation and experiment. We demonstrate two novel approaches to characterize local environments in different polymorphs of molecular crystals using learning models that employ either flexibly learned or handcrafted molecular representations. In the first case, we follow our earlier work on graph learning in molecular crystals, deploying an atomistic graph convolutional network, combined with molecule-wise aggregation, to enable per-molecule environmental classification. For the second model, we develop a new set of descriptors based on symmetry functions combined with a point-vector representation of the molecules, encoding information about the positions as well as relative orientations of the molecule. We demonstrate very high classification accuracy for both approaches on urea and nicotinamide crystal polymorphs, and practical applications to the analysis of dynamical trajectory data for nanocrystals and solid-solid interfaces. Both architectures are applicable to a wide range of molecules and diverse topologies, providing an essential step in the exploration of complex condensed matter phenomena.

In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.

We develop and test new machine learning strategies for accelerating molecular crystal structure ranking and crystal property prediction using tools from geometric deep learning on molecular graphs. Leveraging developments in graph-based learning and the availability of large molecular crystal datasets, we train models for density prediction and stability ranking which are accurate, fast to evaluate, and applicable to molecules of widely varying size and composition. Our density prediction model, MolXtalNet-D, achieves state of the art performance, with lower than 2% mean absolute error on a large and diverse test dataset. Our crystal ranking tool, MolXtalNet-S, correctly discriminates experimental samples from synthetically generated fakes and is further validated through analysis of the submissions to the Cambridge Structural Database Blind Tests 5 and 6. Our new tools are computationally cheap and flexible enough to be deployed within an existing crystal structure prediction pipeline both to reduce the search space and score/filter crystal candidates.

Winning the coin toss at the end of a tied soccer game gives a team the right to choose whether to kick either first or second on all five rounds of penalty kicks, when each team is allowed one kick per round. There is considerable evidence that the right to make this choice, which is usually to kick first, gives a team a significant advantage. To make the outcome of a tied game fairer, we suggest a rule that handicaps the team that kicks first (A), requiring it to succeed on one more penalty kick than the team that kicks second (B). We call this the $m - n$ rule and, more specifically, propose $(m, n)$ = (5, 4): For A to win, it must successfully kick 5 goals before the end of the round in which B kicks its 4th; for B to win, it must succeed on 4 penalty kicks before A succeeds on 5. If both teams reach (5, 4) on the same round -- when they both kick successfully at (4, 3) -- then the game is decided by round-by-round "sudden death," whereby the winner is the first team to score in a subsequent round when the other team does not. We show that this rule is fair in tending to equalize the ability of each team to win a tied game in a penalty shootout. We also discuss a related rule that precludes the teams from reaching (5, 4) at the same time, obviating the need for sudden death and extra rounds.

Disordered molecular systems such as amorphous catalysts, organic thin films, electrolyte solutions, and water are at the cutting edge of computational exploration today. Traditional simulations of such systems at length-scales relevant to experiments in practice require a compromise between model accuracy and quality of sampling. To remedy the situation, we have developed an approach based on generative machine learning called the Morphological Autoregressive Protocol (MAP) which provides computational access to mesoscale disordered molecular configurations at linear cost at generation for materials in which structural correlations decay sufficiently rapidly. The algorithm is implemented using an augmented PixelCNN deep learning architecture that we previously demonstrated produces excellent results in 2 dimensions (2D) for mono-elemental molecular systems. Here, we extend our implementation to multielemental 3D and demonstrate performance using water as our test system in two scenarios: 1. liquid water, and 2. a sample conditioned on the presence of a rare motif. We trained the model on small-scale samples of liquid water produced using path-integral molecular dynamics simulation including nuclear quantum effects under ambient conditions. MAP-generated water configurations are shown to accurately reproduce the properties of the training set and to produce stable trajectories when used as initial conditions in classical and quantum dynamical simulations. We expect our approach to perform equally well on other disordered molecular systems while offering unique advantages in situations when the disorder is quenched rather than equilibrated.

Inverse design of short single-stranded RNA and DNA sequences (aptamers) is the task of finding sequences that satisfy a set of desired criteria. Relevant criteria may be, for example, the presence of specific folding motifs, binding to molecular ligands, sensing properties, etc. Most practical approaches to aptamer design identify a small set of promising candidate sequences using high-throughput experiments (e.g. SELEX), and then optimize performance by introducing only minor modifications to the empirically found candidates. Sequences that possess the desired properties but differ drastically in chemical composition will add diversity to the search space and facilitate the discovery of useful nucleic acid aptamers. Systematic diversification protocols are needed. Here we propose to use an unsupervised machine learning model known as the Potts model to discover new, useful sequences with controllable sequence diversity. We start by training a Potts model using the maximum entropy principle on a small set of empirically identified sequences unified by a common feature. To generate new candidate sequences with a controllable degree of diversity, we take advantage of the model's spectral feature: an energy bandgap separating sequences that are similar to the training set from those that are distinct. By controlling the Potts energy range that is sampled, we generate sequences that are distinct from the training set yet still likely to have the encoded features. To demonstrate performance, we apply our approach to design diverse pools of sequences with specified secondary structure motifs in 30-mer RNA and DNA aptamers.

Cytokine release syndrome (CRS), also known as cytokine storm, is one of the most consequential adverse effects of chimeric antigen receptor therapies that have shown promising results in cancer treatment. When emerging, CRS could be identified by the analysis of specific cytokine and chemokine profiles that tend to exhibit similarities across patients. In this paper, we exploit these similarities using machine learning algorithms and set out to pioneer a meta--review informed method for the identification of CRS based on specific cytokine peak concentrations and evidence from previous clinical studies. We argue that such methods could support clinicians in analyzing suspect cytokine profiles by matching them against CRS knowledge from past clinical studies, with the ultimate aim of swift CRS diagnosis. During evaluation with real--world CRS clinical data, we emphasize the potential of our proposed method of producing interpretable results, in addition to being effective in identifying the onset of cytokine storm.

We propose a method of separating a desired sound source from a single-channel mixture, based on either a textual description or a short audio sample of the target source. This is achieved by combining two distinct models. The first model, SoundWords, is trained to jointly embed both an audio clip and its textual description to the same embedding in a shared representation. The second model, SoundFilter, takes a mixed source audio clip as an input and separates it based on a conditioning vector from the shared text-audio representation defined by SoundWords, making the model agnostic to the conditioning modality. Evaluating on multiple datasets, we show that our approach can achieve an SI-SDR of 9.1 dB for mixtures of two arbitrary sounds when conditioned on text and 10.1 dB when conditioned on audio. We also show that SoundWords is effective at learning co-embeddings and that our multi-modal training approach improves the performance of SoundFilter., Comment: Submitted to INTERSPEECH 2022

Design of de novo biological sequences with desired properties, like protein and DNA sequences, often involves an active loop with several rounds of molecule ideation and expensive wet-lab evaluations. These experiments can consist of multiple stages, with increasing levels of precision and cost of evaluation, where candidates are filtered. This makes the diversity of proposed candidates a key consideration in the ideation phase. In this work, we propose an active learning algorithm leveraging epistemic uncertainty estimation and the recently proposed GFlowNets as a generator of diverse candidate solutions, with the objective to obtain a diverse batch of useful (as defined by some utility function, for example, the predicted anti-microbial activity of a peptide) and informative candidates after each round. We also propose a scheme to incorporate existing labeled datasets of candidates, in addition to a reward function, to speed up learning in GFlowNets. We present empirical results on several biological sequence design tasks, and we find that our method generates more diverse and novel batches with high scoring candidates compared to existing approaches., Comment: ICML 2022. 15 pages, 3 figures. Code available at: https://github.com/MJ10/BioSeq-GFN-AL. Updated GFP results

In molecular electronic conduction, exotic lattice morphologies often give rise to exotic behaviors. Among 2D systems, graphene is a notable example. Recently, a stable amorphous version of graphene called Monolayer Amorphous Carbon (MAC) was synthesized. MAC poses a new set of questions regarding the effects of disorder on conduction. In this Letter, we perform ensemble-level computational analysis of the coherent electronic transmission through MAC nano-fragments in search of defining characteristics. Our analysis, relying on a semi-empirical Hamiltonian (Pariser-Parr-Pople) and Landauer theory, showed that states near the Fermi energy ($E_F$) in MAC inherit partial characteristics of analogous surface states in graphene nano-fragments. Away from $E_F$, current is carried by a set of delocalized states which transition into a subset of insulating interior states at the band edges. Finally, we also found that quantum interference between the frontier orbitals is a common feature among MAC nano-fragments.

Learning to recognize new keywords with just a few examples is essential for personalizing keyword spotting (KWS) models to a user's choice of keywords. However, modern KWS models are typically trained on large datasets and restricted to a small vocabulary of keywords, limiting their transferability to a broad range of unseen keywords. Towards easily customizable KWS models, we present KeySEM (Keyword Speech EMbedding), a speech embedding model pre-trained on the task of recognizing a large number of keywords. Speech representations offered by KeySEM are highly effective for learning new keywords from a limited number of examples. Comparisons with a diverse range of related work across several datasets show that our method achieves consistently superior performance with fewer training examples. Although KeySEM was pre-trained only on English utterances, the performance gains also extend to datasets from four other languages indicating that KeySEM learns useful representations well aligned with the task of keyword spotting. Finally, we demonstrate KeySEM's ability to learn new keywords sequentially without requiring to re-train on previously learned keywords. Our experimental observations suggest that KeySEM is well suited to on-device environments where post-deployment learning and ease of customization are often desirable., Comment: In INTERSPEECH 2021

User studies have shown that reducing the latency of our simultaneous lecture translation system should be the most important goal. We therefore have worked on several techniques for reducing the latency for both components, the automatic speech recognition and the speech translation module. Since the commonly used commitment latency is not appropriate in our case of continuous stream decoding, we focused on word latency. We used it to analyze the performance of our current system and to identify opportunities for improvements. In order to minimize the latency we combined run-on decoding with a technique for identifying stable partial hypotheses when stream decoding and a protocol for dynamic output update that allows to revise the most recent parts of the transcription. This combination reduces the latency at word level, where the words are final and will never be updated again in the future, from 18.1s to 1.1s without sacrificing performance in terms of word error rate.

With the rise of low power speech-enabled devices, there is a growing demand to quickly produce models for recognizing arbitrary sets of keywords. As with many machine learning tasks, one of the most challenging parts in the model creation process is obtaining a sufficient amount of training data. In this paper, we explore the effectiveness of synthesized speech data in training small, spoken term detection models of around 400k parameters. Instead of training such models directly on the audio or low level features such as MFCCs, we use a pre-trained speech embedding model trained to extract useful features for keyword spotting models. Using this speech embedding, we show that a model which detects 10 keywords when trained on only synthetic speech is equivalent to a model trained on over 500 real examples. We also show that a model without our speech embeddings would need to be trained on over 4000 real examples to reach the same accuracy.

We propose the Fr\'echet Audio Distance (FAD), a novel, reference-free evaluation metric for music enhancement algorithms. We demonstrate how typical evaluation metrics for speech enhancement and blind source separation can fail to accurately measure the perceived effect of a wide variety of distortions. As an alternative, we propose adapting the Fr\'echet Inception Distance (FID) metric used to evaluate generative image models to the audio domain. FAD is validated using a wide variety of artificial distortions and is compared to the signal based metrics signal to distortion ratio (SDR), cosine distance and magnitude L2 distance. We show that, with a correlation coefficient of 0.52, FAD correlates more closely with human perception than either SDR, cosine distance or magnitude L2 distance, with correlation coefficients of 0.39, -0.15 and -0.01 respectively.

We develop and test a computational framework to study heat exchange in interacting, nonequilibrium open quantum systems. Our iterative full counting statistics path integral (iFCSPI) approach extends a previously well-established influence functional path integral method, by going beyond reduced system dynamics to provide the cumulant generating function of heat exchange. The method is straightforward; we implement it for the nonequilibrium spin boson model to calculate transient and long-time observables, focusing on the steady-state heat current flowing through the system under a temperature difference. Results are compared to perturbative treatments and demonstrate good agreement in the appropriate limits. The challenge of converging nonequilibrium quantities, currents and high order cumulants, is discussed in detail. The iFCSPI, a numerically exact technique, naturally captures strong system-bath coupling and non-Markovian effects of the environment. As such, it is a promising tool for probing fundamental questions in quantum transport and quantum thermodynamics.

Low power digital signal processors (DSPs) typically have a very limited amount of memory in which to cache data. In this paper we develop efficient bottleneck feature (BNF) extractors that can be run on a DSP, and retrain a baseline large-vocabulary continuous speech recognition (LVCSR) system to use these BNFs with only a minimal loss of accuracy. The small BNFs allow the DSP chip to cache more audio features while the main application processor is suspended, thereby reducing the overall battery usage. Our presented system is able to reduce the footprint of standard, fixed point DSP spectral features by a factor of 10 without any loss in word error rate (WER) and by a factor of 64 with only a 5.8% relative increase in WER., Comment: Submitted to ICASSP 2019

Service sports include two-player contests such as volleyball, badminton, and squash. We analyze four rules, including the Standard Rule (SR), in which a player continues to serve until he or she loses. The Catch-Up Rule (CR) gives the serve to the player who has lost the previous point - as opposed to the player who won the previous point, as under SR. We also consider two Trailing Rules that make the server the player who trails in total score. Surprisingly, compared with SR, only CR gives the players the same probability of winning a game while increasing its expected length, thereby making it more competitive and exciting to watch. Unlike one of the Trailing Rules, CR is strategy-proof. By contrast, the rules of tennis fix who serves and when; its tiebreaker, however, keeps play competitive by being fair - not favoring either the player who serves first or who serves second., Comment: 31 pages. Forthcoming in The American Mathematical Monthly, 2018

Absorption refrigerators transfer thermal energy from a cold reservoir to a hot reservoir using input energy from a third, so-called work reservoir. We examine the operation of quantum absorption refrigerators when coherences between eigenstates survive in the steady state limit. In our model, the working medium comprises a discrete, four-level system. We manifest that eigenbasis quantum coherences within this system generally suppress the cooling current in the refrigerator, while minimally affecting the coefficient of performance (cooling efficiency). We rationalize the behavior of the four-level refrigerator by studying two, three-level model systems for energy transport and refrigeration. Our calculations further illuminate the shortcomings of secular quantum master equations, and the necessity of employing dynamical equations of motion that retain couplings between population and coherences.

Existing music recognition applications require a connection to a server that performs the actual recognition. In this paper we present a low-power music recognizer that runs entirely on a mobile device and automatically recognizes music without user interaction. To reduce battery consumption, a small music detector runs continuously on the mobile device's DSP chip and wakes up the main application processor only when it is confident that music is present. Once woken, the recognizer on the application processor is provided with a few seconds of audio which is fingerprinted and compared to the stored fingerprints in the on-device fingerprint database of tens of thousands of songs. Our presented system, Now Playing, has a daily battery usage of less than 1% on average, respects user privacy by running entirely on-device and can passively recognize a wide range of music., Comment: Authors are listed in alphabetical order by last name

We study the electrical conductance $G$ and the thermopower $S$ of single-molecule junctions, and reveal signatures of different transport mechanisms: off-resonant tunneling, on-resonant coherent (ballistic) motion, and multi-step hopping. These mechanisms are identified by studying the behavior of $G$ and $S$ while varying molecular length and temperature. Based on a simple one-dimensional model for molecular junctions, we derive approximate expressions for the thermopower in these different regimes. Analytical results are compared to numerical simulations, performed using a variant of B\"uttiker's probe technique, the so-called voltage-temperature probe, which allows us to phenomenologically introduce environmentally-induced elastic and inelastic electron scattering effects, while applying both voltage and temperature biases across the junction. We further simulate the thermopower of GC-rich DNA molecules with mediating A:T blocks, and manifest the tunneling-to-hopping crossover in both the electrical conductance and the thermopower, in accord with measurements by Y. Li et al., Nature Comm. 7, 11294 (2016).

We study an intermediate quantum coherent-incoherent charge transport mechanism in metal-molecule-metal junctions using B\"uttiker's probe technique. This tool allows us to include incoherent effects in a controlled manner, and thus to study situations in which partial decoherence affects charge transfer dynamics. Motivated by recent experiments on intermediate coherent-incoherent charge conduction in DNA molecules [L. Xiang {\it et al.}, Nature Chem. 7, 221-226 (2015)], we focus on two representative structures: alternating (GC)$_n$ and stacked G$_n$C$_n$ sequences; the latter structure is argued to support charge delocalization within G segments, and thus an intermediate coherent-incoherent conduction. We begin our analysis with a highly simplified 1-dimensional tight-binding model, while introducing environmental effects through B\"uttiker's probes. This minimal model allows us to gain fundamental understanding of transport mechanisms and derive analytic results for molecular resistance in different limits. We then use a more detailed ladder-model Hamiltonian to represent double-stranded DNA structures---with environmental effects captured by B\"uttiker's probes. We find that hopping conduction dominates in alternating sequences, while in stacked sequences charge delocalization (visualized directly through the electronic density matrix) supports significant resonant-ballistic charge dynamics reflected by an even-odd effect and a weak distance dependence for resistance. Our analysis illustrates that lessons learned from minimal models are helpful for interpreting charge dynamics in DNA., Comment: 16 pages, 14 figures

We extend the Landauer-B\"uttiker probe formalism for conductances to the high bias regime, and study the effects of environmentally-induced elastic and inelastic scattering on charge current in single molecule junctions, focusing on high-bias effects. The probe technique phenomenologically incorporates incoherent elastic and inelastic effects to the fully coherent case, mimicking a rich physical environment at trivial cost. We further identify environmentally-induced mechanisms which generate an asymmetry in the current, manifested as a weak diode behavior. This rectifying behavior, found in two types of molecular junction models, is absent in the coherent-elastic limit, and is only active in the case with incoherent-inelastic scattering. Our work illustrates that in the low bias - linear response regime, the commonly used "dephasing probe" (mimicking only elastic decoherence effects) operates nearly indistinguishably from a "voltage probe" (admitting inelastic-dissipative effects). However, these probes realize fundamentally distinct $I$-$V$ characteristics at high biases, reflecting the central roles of dissipation and inelastic scattering processes on molecular electronic transport far-from-equilibrium.

We examine the robustness of single-molecule tunneling diodes to thermal-environmental effects. The diode comprises three fragments: two different conjugated chemical groups at the boundaries, and a saturated moiety in between, breaking conjugation. In this setup, molecular electronic levels localized on the conjugated groups independently shift with applied bias. While in the forward polarity a resonance condition is met, enhancing conductance, in the reversed direction molecular electronic states shift away from each other, resulting in small tunneling currents. In the absence of interactions with a thermal environment (consisting e.g. internal vibrations, solvent), rectification ratios reach three orders of magnitude. We introduce decoherence and inelastic-dissipative effects phenomenologically, by using the "voltage probe" approach. We find that when $\gamma_d \lesssim v$, with $\gamma_d$ the interaction energy of electrons with the environment and $v$ the tunneling energy across the saturated link, the diode is still highly effective, though rectification ratios are cut down by a factor of 2-4 compared to the coherent limit. To further enhance rectification ratios in molecular diodes we suggest a refined design involving four orbitals, with a pair of closely spaced states at each conjugated moiety.

We demonstrate that a simple phenomenological approach can be used to simulate electronic conduction in molecular wires under thermal effects induced by the surrounding environment. This "Landauer-B\"uttiker's probe technique" can properly replicate different transport mechanisms: phase coherent nonresonant tunneling, ballistic behavior, and hopping conduction, to provide results consistent with experiments. Specifically, our simulations with the probe method recover the following central characteristics of charge transfer in molecular wires: (i) The electrical conductance of short wires falls off exponentially with molecular length, a manifestation of the tunneling (superexchange) mechanism. Hopping dynamics overtakes superexchange in long wires demonstrating an ohmic-like behavior. (ii) In off-resonance situations, weak dephasing effects facilitate charge transfer. Under large dephasing the electrical conductance is suppressed. (iii) At high enough temperatures, $k_BT/\epsilon_B>1/25$, with $\epsilon_B$ as the molecular-barrier height, the current is enhanced by a thermal activation (Arrhenius) factor. However, this enhancement takes place for both coherent and incoherent electrons and it does not readily indicate the underlying mechanism. (iv) At finite-bias, dephasing effects impede conduction in resonant situations. We further show that memory (non-Markovian) effects can be implemented within the Landauer-B\"uttiker's probe technique to model the interaction of electrons with a structured environment. Finally, we examine experimental results of electron transfer in conjugated molecular wires and show that our computational approach can reasonably reproduce reported values to provide mechanistic information.

Argues that new patterns of trade and production combined with an emphasis on a knowledge-based economy/society make it imperative that Canada upgrade its educational system. Specifically notes that several growing and dominant industries (microelectronics, biotechnology, telecommunications) require a high-tech skilled labor force. (MJP)

The percentage of books in a research library having personal authors that were cataloged and entered in a one-screen minicatalog was determined to be 36.6 percent when searched. Libraries can reduce cataloging costs and increase user access to information by simplifying cataloging. (AEF)