Mark M. Wilde

I set up an account on blue sky now and will be actively posting there. It seems most of the q. info. community has left twitter for the moment. I'll still log on to twitter to check and see what people are posting here, but I won't post anything here for now

@MvsCerezo is it a typo? did you mean tilde instead of wilde? My little joke with sandwiched Renyi relative entropy is that the tilde is for wilde :)

@MvsCerezo you have a LaTeX command for me? :)

@markwilde randomly pops up in my overleaf 👀

@EricAspling @QuantumMemeing now using this for an upcoming talk in Korea :)

this looks like the kind of textbook that can only be found in a Soviet Era library in Moscow, and i love it

@maria_demidik @cenk_tuysuz @GrosQmichi @desy_cqta @CERNquantum ok, thanks for your reply.

@markwilde @cenk_tuysuz @GrosQmichi @desy_cqta @CERNquantum Yes, you are definitely right! We wanted to rewrite it explicitly to point to the problem of computing the positive phase. We in fact cite this work multiple times throughout the paper. We will revise our wording to make this more clear. Thank you for your comment.

📢 A trainable alternative to circuit-based QML: semi-quantum Restricted Boltzmann machines (sqRBM) Recently, we have witnessed lots of results connected to trainability and scalability issues of circuit-based models 😤 Is there an alternative way to design QML models? 🧵👇

@maria_demidik @cenk_tuysuz @GrosQmichi @desy_cqta @CERNquantum wasn't your Prop. 1 already shown in Eq. (18) of the original QBM paper arxiv.org/abs/1601.02036 ? (It is hard to see any different between your Prop. 1 and Eq. (18) of their paper, and no reference or discussion of this point has been given.)

Many thanks to my collaborators! @cenk_tuysuz @GrosQmichi, Nico Piatkowski and Karl Jansen, @desy_cqta @CERNquantum

@FarzinSalek @Perimeter thanks! I really wish this talk had been about hockey-stick divergences and will do that next time

When in Canada, use a hockey stick! @markwilde delivers his talk on quantum hypothesis exclusion at @Perimeter with a true Canadian twist 🏒

slides for talk "Machina Ex Quanta: Rise of the Quantum Boltzmann Machines" now available at zenodo.org/records/146335…

@GuglielmoMazzo3 @MvsCerezo @qZoeHolmes @QuantSciTech this is nothing. I once had a paper desk rejected from Communication in Mathematical Physics after six months. It happened only because I emailed the editor and asked about the status of my paper. Then the editor replied and said "I thought I told you it was desk rejected" :)

@MvsCerezo @qZoeHolmes @QuantSciTech Oh, last time I submitted to this journal, the editors took more than a month for a desk rejection. I think I'll pass :)

🖊️ @qZoeHolmes and I are guest-editing a focus issue for @QuantSciTech journal. We want - (Wacky) new ideas for variational algorithms - Honest benchmarks, if you show it don't work, we want it (be thorough) - Studies on classical simulability - Power of data in QML And more!

#キャルちゃんのquantphチェック Evolved (進化した？) 量子ボルツマンマシンの提案。パラメータ化された2つのハミルトニアンG, Hに対し、Gの熱状態を準備、次にHを用いユニタリ時間発展を行うというもの。これの基底状態エネルギー推定や生成モデリングへの応用を示した。 arxiv.org/abs/2501.03367

we also thank those who wrote helpful reviews on quantum generalizations of Fisher info, without which we would not have been able to make the progress reported in this paper, arxiv.org/abs/2410.24058, & arxiv.org/abs/2411.18268. This includes @sidhu_jasminder @jj_xyz @FarisSbahi

As a bonus, we prove that for general parameterized families of states the Fisher-Bures and Wigner-Yanase information matrices differ by no more than a factor of two in the matrix Loewner order, and we discuss applications in metric-aware gradient descent and estimation tasks

Joint work "Evolved Quantum Boltzmann Machines" w/ Michele Minervini & Dhrumil Patel now online: arxiv.org/abs/2501.03367 Our paper has 3 main contributions: (see below) 1/n

@LamiLudovico @scuolanormale @ERC_Research @UvA_Amsterdam @QuSoftAmsterdam Congrats again!

Settling in at my new desk at my new home, the @scuolanormale in Pisa, Italy. I'm excited to be back to my alma mater and to set up here my own @ERC_Research-funded group soon. Heartfelt thanks to @UvA_Amsterdam and @QuSoftAmsterdam for two unforgettable years!

Oldies but goldies: B Poljak, Some methods of speeding up the convergence of iteration methods, 1964. Introduces momentum methods to speed up gradient descent schemes. distill.pub/2017/momentum/

@Kaju_Nut so then we really want the mathematical proof to have it in general, right?

@markwilde Indeed.

Proofs of strong subadditivity of von Neumann entropy: Direct proof: 4 pages of dense math. Proof via holography: One(1) picture. Holography is magic