New paper 🚨 https://t.co/GWlCpiAsjB “Improved Approximation Algorithms for Low-Rank Problems Using Semidefinite Optimization” (w/ Jean Pauphilet) Inspired by Goemans-Williamson’s success in binary quadratic optimization, we generalize to semi-orthogonal and low-rank matrices.

The Analytics & Operations group @ImperialBiz Imperial Business School is hiring for an open-ranked position (especially interested in rookies). Deadline 8 October, please apply via the link here:

There are now sixteen!! items in the NeurIPS checklist

🚨 The INFORMS Computing Society (ICS) invites submissions for its 𝟮𝟬𝟮𝟱 𝗦𝘁𝘂𝗱𝗲𝗻𝘁 𝗣𝗮𝗽𝗲𝗿 𝗔𝘄𝗮𝗿𝗱, recognizing outstanding research at the interface of computing and operations research by a student author. For details:

We concluded the conference with a dinner at Worcester College which made me feel like I was in The Great Hall of Harry Potter! Looking forward to the conference next year (location to be announced)

Finally, @bodonoghue85 (DeepMind, not pictured) gave an overview of the recent successes of Google DeepMind in terms of using reinforcement learning and generative AI to solve challenging problems like the Mathematics Olympiad as a step towards artificial general intelligence.

Eight, Paul Goulart (Oxford) talked about the Clarabel.jl solver for conic programs, and discussed how formulating quadratic programs in a quadratic way (rather than reformulating them as SOCPs) can improve the performance of an interior point method https://t.co/PHkcoDaKFl

Seventh, Pierre Pinson (Imperial, not pictured) investigated trading strategies for price-taking participants in energy markets as studied in "Distributionally robust trading strategies for renewable energy producers" https://t.co/GQFnUEiKVt

Sixth, my colleague @gahyiban (Imperial College Business School) analyzed a retailer that provides an online sharing economy as analyzed in the working paper "Personalized Assortment Optimization for a Subscription Business Model of Experience Goods" https://t.co/7ACeSzSS7S

Fifth, Neil Walton (Durham) discussed modeling the NHS waitlist as queueing system and analyzed how the UK Labour Party's pledge that most patients should have access to elective procedures within 18 weeks is going (TL:DR progress, not enough to meet pledge within current term) .

Fourth, Stefan Scholtes (Cambridge) talked about market power in drug supply chains as studied in https://t.co/JuVHBDqBge ending with a call for other researchers to also investigate the long-run implications of creating non-for-profit drug manufacturers like Civica Rx

Third, Rouba Ibrahim (UCL) talked about threshold policies for reducing the length of queues in queueing systems as discussed in "The Effects of Information on Abandonment and Congestion in Non-Stationary Priority Queues https://t.co/JZfoBauAUB

Second, Nitish Jain (LBS, not pictured) talked about expiry dates in food waste and policies for mitigating food waste as discussed in "Until Later is Preferred Over Sooner: Multiplicity in Product Expiration Dates and Food Waste in Retail Stores"

First up we had a fascinating talk from Matthias Holweg (Oxford) on the role of generative AI (chat-GPT etc.) in academic research and right-sizing editorial policies

On 4 April, had the pleasure of organizing the 2nd London Operations Research Day with Agni Orfanoudaki, Yi-Chun Akchen and Jean Pauphilet at Oxford. Thx to Qube Research and Technology for sponsorship! A recap thread ⬇️ https://t.co/wRKqIefzGw

In our latest #IBKnowledge article, Professor Eddie Anderson discusses overcoming the risk and uncertainty of the energy sector. ⚡️ 🌍 Eddie tells us why a creative approach to analytics is crucial to managing the energy sector. Read more 🔗 → https://t.co/InZy26jeO2

We are accepting applications for a postdoctoral position at London Business School: https://t.co/GLecPqYyNV

We also develop new convex relaxations for low-rank problems, via Shor relaxations. Our original relaxations involve matrices of size n^2 x n^2, which is not tractable. However, we show how to make our relaxations block decomposable, and thus tractable, for several applications.

We propose techniques for combining single-tree Benders decomposition schemes with sampling that maintain deterministic upper and lower bounds. Optimality gap often smaller than 5-10% in minutes-hours even with 100s of nodes in network design problems!

Delighted to share that our paper “A Stochastic Benders Decomposition Scheme for Large-Scale Stochastic Network Design” (w/ @dbertsim Jean Pauphilet and Periklis Petridis) has been accepted at INFORMS Journal on Computing!