Turkish Mosaic Art

Turkish Mosaic Art

Infinite Factor Authentication

Infinite Factory Authentication is coming back to same point after you complete all sorts of different authentication steps.

Presenting on this Sunday

A Hierarchical Descriptor Framework for On-the-Fly Anatomical Location Matching between Longitudinal Studies

Finetuning

#Finetuning in deep learning can be thought of as adjusting an already trained instrument to suit new data better. Consider a simple example where you have a neural network with just one parameter that's designed to predict people's heights using mean squared error (MSE) loss. Ideally, the best prediction would be the average height. Let's say you train this model on a vast dataset, and now you have this pre-trained model. Intuitively, this network has "seen" many examples and has formed a generalized understanding of average height. Now, imagine introducing a new dataset. If you were to train the model solely on this new data, the model would lose its memory of the old data. The learned average might drastically shift if the new population's height is different from the old one. On the other hand, if you incorporate the knowledge from both datasets, your prediction would likely be somewhere between the averages of the two datasets. From a Bayesian standpoint, the ideal balance would give weights to the old and new datasets proportional to their sizes. So, if the new dataset has more data points, it would influence the prediction more heavily. However, if your primary interest is in understanding the new dataset, you might prioritize it and disregard the old data. The core challenge in fine-tuning lies here: how do you strike a balance between retaining useful knowledge from prior data and adapting to new information? This issue is crucial as models venture into new domains or deal with evolving data streams.

Azure Devops

I have faced such a challenging "I am not a robot" in azure devops page such that it requires us to solve after a few trials. It was like a game.

Visual Studio Code Live Preview

Visual studio code does not allow to look at your servers. But if you iframe them it is nicely showing

Ideas

Maybe unrealistic but I kept thinking about these ideas: 1. Getting carbondioxide from air seperating into carbon and oxigen, and squeezing carbon to 3d print diamond houses. 2. Melting and purifying plastic in passific ocean to create an artificial continent. 3. Using sand to produce glasses in desert by glass lenses to produce self replicating machines

E ink

I am typically curious about e-ink technologies. There could be more applications where they are offering now. For example if rooftops covered with them it could provide low cost climate control system.

GPT

While people are able to write more and more, I am willing to read less and less

Time

I the task is urgent I will use existing API. If I have more time I prefer to develop my own framework. If I have more time I prefer to prove math theorem.

WSL is not able to connect internet

https://stackoverflow.com/questions/62314789/no-internet-connection-on-wsl-ubuntu-windows-subsystem-for-linux

Relu functions

Despite their practical success, Relu activations are theoretically ill-defined. This is due to the existence of numerous trivial local optima where the gradient is zero, and we are optimizing them with approximate local optimizers.

Optimization

#Optimization is the core of #machinelearning. All GPU acceleration, cloud systems, and transfer learning in ML are for better optimization. There are many attempts to make it faster. However, there are no widely adopted second-order methods in the field despite some momentum methods depicting them. I believe there is still potential in this area. Having a robust 10% better algorithm could have billion of dollars of impact on the industry. Of course, it is very hard to achieve.

Deep Generative Neural Embeddings for High Dimensional Data Visualization

https://arxiv.org/pdf/2302.10801.pdf

Smile Pass Experiment Started

Smilepass is face-tracking software, as you will see on this page. You can try it yourself. First, you need to add pictures using either taking pictures by the capture button or loading multiple images from a folder. When it is ready, you can use the pass button to continue scanning faces. You can download the pass list once you need it. This software will help to track attendance or eliminate the need for tickets or RFID cards in your organization. Please contact us for any customization or requests. https://lesson.land/smilepass/

Cloud leak

Cloud leak is like memory leak, your bill grows over time while your intent remains the same

Apples and oranges

Where to use ai is more important that having ai inside. If you design a classifier seperating apples and oranges you should first recognize they do not grow in the same tree.

Web2 Web3

Layoffs from cloud companies Amazon, Google and Microsoft is fade of Web2. But Metaverse is failed attempt to create Web3 so far. My perspective on Web3 is on the edge, decentralized data storages. It respects the privacy of the data and give ownership back to users. It requires different set of business models instead of cloud computing and advertisement revenues. Instead, user should be able to list their own items in marketplaces and marketplace owner taxes a small amount for transactions. It allows users to create their own centralized web presence while systems working decentralized.

Tech Needed

The most important need nowadays, in my perspective, is advancement in battery technology. I have soo many devices that I need to charge. All have different interfaces and short life spans. #technology

Start the year with language advice

Java is in continuous decline. Avoid as much as possible. The must known languages are python, javascript and C++