-
A Better Match for Drivers and Riders: Reinforcement Learning at Lyft
Authors:
Xabi Azagirre,
Akshay Balwally,
Guillaume Candeli,
Nicholas Chamandy,
Benjamin Han,
Alona King,
Hyungjun Lee,
Martin Loncaric,
Sebastien Martin,
Vijay Narasiman,
Zhiwei,
Qin,
Baptiste Richard,
Sara Smoot,
Sean Taylor,
Garrett van Ryzin,
Di Wu,
Fei Yu,
Alex Zamoshchin
Abstract:
To better match drivers to riders in our ridesharing application, we revised Lyft's core matching algorithm. We use a novel online reinforcement learning approach that estimates the future earnings of drivers in real time and use this information to find more efficient matches. This change was the first documented implementation of a ridesharing matching algorithm that can learn and improve in rea…
▽ More
To better match drivers to riders in our ridesharing application, we revised Lyft's core matching algorithm. We use a novel online reinforcement learning approach that estimates the future earnings of drivers in real time and use this information to find more efficient matches. This change was the first documented implementation of a ridesharing matching algorithm that can learn and improve in real time. We evaluated the new approach during weeks of switchback experimentation in most Lyft markets, and estimated how it benefited drivers, riders, and the platform. In particular, it enabled our drivers to serve millions of additional riders each year, leading to more than $30 million per year in incremental revenue. Lyft rolled out the algorithm globally in 2021.
△ Less
Submitted 13 November, 2023; v1 submitted 20 October, 2023;
originally announced October 2023.
-
Unconditionally secure digital signatures implemented in an 8-user quantum network
Authors:
Yoann Pelet,
Ittoop Vergheese Puthoor,
Natarajan Venkatachalam,
Sören Wengerowsky,
Martin Lončarić,
Sebastian Philipp Neumann,
Bo Liu,
Željko Samec,
Mario Stipčević,
Rupert Ursin,
Erika Andersson,
John G. Rarity,
Djeylan Aktas,
Siddarth Koduru Joshi
Abstract:
The ability to know and verifiably demonstrate the origins of messages can often be as important as encrypting the message itself. Here we present an experimental demonstration of an unconditionally secure digital signature (USS) protocol implemented for the first time, to the best of our knowledge, on a fully connected quantum network without trusted nodes. Our USS protocol is secure against forg…
▽ More
The ability to know and verifiably demonstrate the origins of messages can often be as important as encrypting the message itself. Here we present an experimental demonstration of an unconditionally secure digital signature (USS) protocol implemented for the first time, to the best of our knowledge, on a fully connected quantum network without trusted nodes. Our USS protocol is secure against forging, repudiation and messages are transferrable. We show the feasibility of unconditionally secure signatures using only bi-partite entangled states distributed throughout the network and experimentally evaluate the performance of the protocol in real world scenarios with varying message lengths.
△ Less
Submitted 10 February, 2022; v1 submitted 9 February, 2022;
originally announced February 2022.
-
Experimental implementation of secure anonymous protocols on an eight-user quantum network
Authors:
Zixin Huang,
Siddarth Koduru Joshi,
Djeylan Aktas,
Cosmo Lupo,
Armanda O. Quintavalle,
Natarajan Venkatachalam,
Sören Wengerowsky,
Martin Lončarić,
Sebastian Philipp Neumann,
Bo Liu,
Željko Samec,
Laurent Kling,
Mario Stipčević,
Rupert Ursin,
John G. Rarity
Abstract:
Anonymity in networked communication is vital for many privacy-preserving tasks. Secure key distribution alone is insufficient for high-security communications, often knowing who transmits a message to whom and when must also be kept hidden from an adversary. Here we experimentally demonstrate 5 information-theoretically secure anonymity protocols on an 8 user city-wide quantum network using polar…
▽ More
Anonymity in networked communication is vital for many privacy-preserving tasks. Secure key distribution alone is insufficient for high-security communications, often knowing who transmits a message to whom and when must also be kept hidden from an adversary. Here we experimentally demonstrate 5 information-theoretically secure anonymity protocols on an 8 user city-wide quantum network using polarisation-entangled photon pairs. At the heart of these protocols is anonymous broadcasting, which is a cryptographic primitive that allows one user to reveal one bit of information while kee** her identity anonymous. For a network of $n$ users, the protocols retain anonymity for the sender, given less than $n-2$ users are dishonest. This is one of the earliest implementations of genuine multi-user cryptographic protocols beyond standard QKD. Our anonymous protocols enhance the functionality of any fully-connected Quantum Key Distribution network without trusted nodes.
△ Less
Submitted 18 November, 2020;
originally announced November 2020.
-
Learning Hash Codes via Hamming Distance Targets
Authors:
Martin Loncaric,
Bowei Liu,
Ryan Weber
Abstract:
We present a powerful new loss function and training scheme for learning binary hash codes with any differentiable model and similarity function. Our loss function improves over prior methods by using log likelihood loss on top of an accurate approximation for the probability that two inputs fall within a Hamming distance target. Our novel training scheme obtains a good estimate of the true gradie…
▽ More
We present a powerful new loss function and training scheme for learning binary hash codes with any differentiable model and similarity function. Our loss function improves over prior methods by using log likelihood loss on top of an accurate approximation for the probability that two inputs fall within a Hamming distance target. Our novel training scheme obtains a good estimate of the true gradient by better sampling inputs and evaluating loss terms between all pairs of inputs in each minibatch. To fully leverage the resulting hashes, we use multi-indexing. We demonstrate that these techniques provide large improvements to a similarity search tasks. We report the best results to date on competitive information retrieval tasks for ImageNet and SIFT 1M, improving MAP from 73% to 84% and reducing query cost by a factor of 2-8, respectively.
△ Less
Submitted 1 October, 2018;
originally announced October 2018.
-
Convolutional Hashing for Automated Scene Matching
Authors:
Martin Loncaric,
Bowei Liu,
Ryan Weber
Abstract:
We present a powerful new loss function and training scheme for learning binary hash functions. In particular, we demonstrate our method by creating for the first time a neural network that outperforms state-of-the-art Haar wavelets and color layout descriptors at the task of automated scene matching. By accurately relating distance on the manifold of network outputs to distance in Hamming space,…
▽ More
We present a powerful new loss function and training scheme for learning binary hash functions. In particular, we demonstrate our method by creating for the first time a neural network that outperforms state-of-the-art Haar wavelets and color layout descriptors at the task of automated scene matching. By accurately relating distance on the manifold of network outputs to distance in Hamming space, we achieve a 100-fold reduction in nontrivial false positive rate and significantly higher true positive rate. We expect our insights to provide large wins for hashing models applied to other information retrieval hashing tasks as well.
△ Less
Submitted 8 February, 2018;
originally announced February 2018.
