Interesting layer skip architecture thing.
Apparently destroyed ImageNet 2015 and became very very famous as such.
- torchvision ResNet
- MLperf v2.1 ResNet contains a pre-trained ResNet ONNX at zenodo.org/record/4735647/files/resnet50_v1.onnx for its inference benchmark. We've tested it at: Run MLperf v2.1 ResNet on Imagenette.
catalog.ngc.nvidia.com/orgs/nvidia/resources/resnet_50_v1_5_for_pytorch explains:
The difference between v1 and v1.5 is that, in the bottleneck blocks which requires downsampling, v1 has stride = 2 in the first 1x1 convolution, whereas v1.5 has stride = 2 in the 3x3 convolution.This difference makes ResNet50 v1.5 slightly more accurate (~0.5% top1) than v1, but comes with a small performance drawback (~5% imgs/sec).
As of 2021, their location is a small business park in Haywards Heath, about 15 minutes north of Brighton[ref]
Funding rounds:
- 2022:
- 67m euro contract with the German government: www.uktech.news/deep-tech/universal-quantum-german-contract-20221102 Both co-founders are German. They then immediatly announced several jobs in Hamburg: apply.workable.com/universalquantum/?lng=en#jobs so presumably linked to the Hamburg University of Technology campus of the German Aerospace Center.
- medium.com/@universalquantum/universal-quantum-wins-67m-contract-to-build-the-fully-scalable-trapped-ion-quantum-computer-16eba31b869e
- 2021: $10M (7.5M GBP) grant from the British Government: www.uktech.news/news/brighton-universal-quantum-wins-grant-20211105This grant is very secretive, very hard to find any other information about it! Most investment trackers are not listing it.The article reads:Interesting!
Universal Quantum will lead a consortium that includes Rolls-Royce, quantum developer Riverlane, and world-class researchers from Imperial College London and The University of Sussex, among others.
A but further down the article gives some more information of partners, from which some of the hardware vendors can be deduced:The consortium includes end-user Rolls-Royce supported by the Science and Technology Facilities Council (STFC) Hartree Centre, quantum software developer Riverlane, supply chain partners Edwards, TMD Technologies (now acquired by Communications & Power Industries (CPI)) and Diamond Microwave
- Edwards is presumably Edwards Vacuum, since we know that trapped ion quantum computers rely heavily on good vacuum systems. Edwards Vacuum is also located quite close to Universal Quantum as of 2022, a few minutes drive.
- TMD Technologies is a microwave technology vendor amongst other things, and we know that microwaves are used e.g. to initialize the spin states of the ions
- Diamond Microwave is another microwave stuff vendor
www.riverlane.com/news/2021/12/riverlane-joins-7-5-million-consortium-to-build-error-corrected-quantum-processor/ gives some more details on the use case provided by Rolls Royce:The work with Rolls Royce will explore how quantum computers can develop practical applications toward the development of more sustainable and efficient jet engines.This starts by applying quantum algorithms to take steps to toward a greater understanding of how liquids and gases flow, a field known as 'fluid dynamics'. Simulating such flows accurately is beyond the computational capacity of even the most powerful classical computers today.This funding was part of a larger quantum push by the UKNQTP: www.ukri.org/news/50-million-in-funding-for-uk-quantum-industrial-projects/ - 2020: $4.5M (3.5M GBP) www.crunchbase.com/organization/universal-quantum. Just out of stealth.
Co-founders:
- Sebastian Weidt. He is German, right? Yes at youtu.be/SwHaJXVYIeI?t=1078 from Video 3. "Fireside Chat with with Sebastian Weidt by Startup Grind Brighton (2022)". The company was founded by two Germans from Essex!
- Winfried Hensinger: if you saw him on the street, you'd think he plays in a punk-rock band. That West Berlin feeling.
Homepage points to foundational paper: www.science.org/doi/10.1126/sciadv.1601540
Universal Quantum emerges out of stealth by University of Sussex (2020)
Source. Explains that a more "traditional" trapped ion quantum computer would user "pairs of lasers", which would require a lot of lasers. Their approach is to try and do it by applying voltages to a microchip instead.- youtu.be/rYe9TXz35B8?t=127 shows some 3D models. It shows how piezoelectric actuators are used to align or misalign some plates, which presumably then determine conductivity
Quantum Computing webinar with Sebastian Weidt by Green Lemon Company (2020)
Source. The sound quality is to bad to stop and listen to, but it presumaby shows the coding office in the background.Fireside Chat with with Sebastian Weidt by Startup Grind Brighton (2022)
Source. Very basic target audience:- youtu.be/SwHaJXVYIeI?t=680 we are not at a point where you can buy victory. There is too much uncertainty involved across different approaches.
- youtu.be/SwHaJXVYIeI?t=949 his background
- youtu.be/SwHaJXVYIeI?t=1277 difference between venture capitalists in different countries
- youtu.be/SwHaJXVYIeI?t=1535 they are 33 people now. They've just setup their office in Haywards Heath, north of Bristol.
By default, the setup runs on CPU only, not GPU, as could be seen by running htop. But by the magic of PyTorch, modifying the program to run on the GPU is trivial:and leads to a faster runtime, with less
cat << EOF | patch
diff --git a/run.py b/run.py
index 104d363..20072d1 100644
--- a/run.py
+++ b/run.py
@@ -24,7 +24,8 @@ data_test = MNIST('./data/mnist',
data_train_loader = DataLoader(data_train, batch_size=256, shuffle=True, num_workers=8)
data_test_loader = DataLoader(data_test, batch_size=1024, num_workers=8)
-net = LeNet5()
+device = 'cuda'
+net = LeNet5().to(device)
criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(net.parameters(), lr=2e-3)
@@ -43,6 +44,8 @@ def train(epoch):
net.train()
loss_list, batch_list = [], []
for i, (images, labels) in enumerate(data_train_loader):
+ labels = labels.to(device)
+ images = images.to(device)
optimizer.zero_grad()
output = net(images)
@@ -71,6 +74,8 @@ def test():
total_correct = 0
avg_loss = 0.0
for i, (images, labels) in enumerate(data_test_loader):
+ labels = labels.to(device)
+ images = images.to(device)
output = net(images)
avg_loss += criterion(output, labels).sum()
pred = output.detach().max(1)[1]
@@ -84,7 +89,7 @@ def train_and_test(epoch):
train(epoch)
test()
- dummy_input = torch.randn(1, 1, 32, 32, requires_grad=True)
+ dummy_input = torch.randn(1, 1, 32, 32, requires_grad=True).to(device)
torch.onnx.export(net, dummy_input, "lenet.onnx")
onnx_model = onnx.load("lenet.onnx")
EOFuser as now we are spending more time on the GPU than CPU:real 1m27.829s
user 4m37.266s
sys 0m27.562s"Magic star" can refer to several different concepts, depending on the context. Here are a few possible interpretations: 1. **Mathematics**: In number theory, a magic star is similar to a magic square. It consists of points arranged in a star shape where the sums of the numbers in a specific formation (such as lines or diagonals) all yield the same total.
Bdellium is a gum resin that is obtained from certain trees in the genus Commiphora, which are part of the Burseraceae family. The term "bdellium" is sometimes used to refer to a resin similar to myrrh. Historically, bdellium has been mentioned in ancient texts, including the Bible, where it is described as a valuable substance. The resin has been used for various purposes, including as an ingredient in perfumes, incense, and traditional medicine.
Ciro Santilli once visited the chemistry department of a world leading university, and the chemists there were obsessed with NMR. They had small benchtop NMR machines. They had larger machines. They had a room full of huge machines. They had them in corridors and on desk tops. Chemists really love that stuff. More precisely, these are used for NMR spectroscopy, which helps identify what a sample is made of.
Introduction to NMR by Allery Chemistry
. Source. - only works with an odd number of nucleons
- apply strong magnetic field, this separates the energy of up and down spins. Most spins align with field.
- send radio waves into sample to make nucleons go to upper energy level. We can see that the energy difference is small since we are talking about radio waves, low frequency.
- when nucleon goes back down, it re-emits radio waves, and we detect that. TODO: how do we not get that confused with the input wave, which is presumably at the same frequency? It appears to send pulses, and then wait for the response.
How to Prepare and Run a NMR Sample by University of Bath (2017)
Source. This is a more direct howto, cool to see. Uses a Bruker Corporation 300. They have a robotic arm add-on. Shows spectrum on computer screen at the end. Shame no molecule identification after that!This video has the merit of showing real equipment usage, including sample preparation.
Says clearly that NMR is the most important way to identify organic compounds.
- youtu.be/uNM801B9Y84?t=41 lists some of the most common targets, including hydrogen and carbon-13
- youtu.be/uNM801B9Y84?t=124 ethanol example
- youtu.be/uNM801B9Y84?t=251 they use solvents where all protium is replaced by deuterium to not affect results. Genius.
- youtu.be/uNM801B9Y84?t=354 usually they do 16 radio wave pulses
Introductory NMR & MRI: Video 01 by Magritek (2009)
Source. Precession and Resonance. Precession has a natural frequency for any angle of the wheel.Introductory NMR & MRI: Video 02 by Magritek (2009)
Source. The influence of temperature on spin statistics. At 300K, the number of up and down spins are very similar. As you reduce temperature, we get more and more on lower energy state.Introductory NMR & MRI: Video 03 by Magritek (2009)
Source. The influence of temperature on spin statistics. At 300K, the number of up and down spins are very similar. As you reduce temperature, we get more and more on lower energy state.NMR spectroscopy visualized by ScienceSketch
. Source. 2020. Decent explanation with animation. Could go into a bit more numbers, but OK.Often used as a synonym for X-ray crystallography, or to refer more specifically to the diffraction part of the experiment (exluding therefore sample preparation and data processing).
Pinned article: Introduction to the OurBigBook Project
Welcome to the OurBigBook Project! Our goal is to create the perfect publishing platform for STEM subjects, and get university-level students to write the best free STEM tutorials ever.
Everyone is welcome to create an account and play with the site: ourbigbook.com/go/register. We belive that students themselves can write amazing tutorials, but teachers are welcome too. You can write about anything you want, it doesn't have to be STEM or even educational. Silly test content is very welcome and you won't be penalized in any way. Just keep it legal!
Intro to OurBigBook
. Source. We have two killer features:
- topics: topics group articles by different users with the same title, e.g. here is the topic for the "Fundamental Theorem of Calculus" ourbigbook.com/go/topic/fundamental-theorem-of-calculusArticles of different users are sorted by upvote within each article page. This feature is a bit like:
- a Wikipedia where each user can have their own version of each article
- a Q&A website like Stack Overflow, where multiple people can give their views on a given topic, and the best ones are sorted by upvote. Except you don't need to wait for someone to ask first, and any topic goes, no matter how narrow or broad
This feature makes it possible for readers to find better explanations of any topic created by other writers. And it allows writers to create an explanation in a place that readers might actually find it.
Figure 1. Screenshot of the "Derivative" topic page. View it live at: ourbigbook.com/go/topic/derivativeVideo 2. OurBigBook Web topics demo. Source. - local editing: you can store all your personal knowledge base content locally in a plaintext markup format that can be edited locally and published either:This way you can be sure that even if OurBigBook.com were to go down one day (which we have no plans to do as it is quite cheap to host!), your content will still be perfectly readable as a static site.
- to OurBigBook.com to get awesome multi-user features like topics and likes
- as HTML files to a static website, which you can host yourself for free on many external providers like GitHub Pages, and remain in full control
Figure 2. You can publish local OurBigBook lightweight markup files to either OurBigBook.com or as a static website.
Figure 3. Visual Studio Code extension installation.
Figure 5. . You can also edit articles on the Web editor without installing anything locally. Video 3. Edit locally and publish demo. Source. This shows editing OurBigBook Markup and publishing it using the Visual Studio Code extension. 
- Infinitely deep tables of contents:
All our software is open source and hosted at: github.com/ourbigbook/ourbigbook
Further documentation can be found at: docs.ourbigbook.com
Feel free to reach our to us for any help or suggestions: docs.ourbigbook.com/#contact