AIModel-Mutator: Finding Vulnerabilities in TensorFlow

Another current study on the security of machine learning models, and information on how framework bugs (such as Tensorflow) can affect it. For example, from 2019 to 2021, the number of CVEs for TF increased 15 times.

Qian Feng, a senior security researcher at Baidu Security talks about the important work they did with their colleagues.

As we know, it's pretty easy to corrupt models, they freely distributed and without any additional checks, so short deep dive into the problem in this video: https://www.youtube.com/watch?v=7QqbJRZ6CxU.

Prepared by @codemining.

GLIDE for image augmentation aka ToadVerse technical details

Technical details on how we used GLIDE for image augmentation.

Article Link: https://mirror.xyz/kefirski.eth/XN1cV27uHcAjN_tPSc_ckgSz4B3Nfh5l5HH9lRs9xEE

#GAN #StyleGAN2 #GLIDE #art #art_generation

​​Simple book about #ML — Machine Learning Simplified

The main purpose of the book is to build an intuitive understanding of how algorithms work through basic examples. In order to understand the presented material, it is enough to know basic mathematics and linear algebra.

After reading this book, you will know the basics of supervised learning, understand complex mathematical models, understand the entire pipeline of a typical ML project, and also be able to share your knowledge with colleagues from related industries and with technical professionals.

And for those who find the theoretical part not enough - the book is supplemented with a repository on GitHub, which has Python implementation of all methods and algorithms described in chapters.

Book is absolutely free to read.

Link: themlsbook.com

#wheretostart #book

AlphaCode Explained: AI Code Generation

AlphaCode is DeepMind's new massive language model for generating code. It is similar to OpenAI Codex, except for in the paper they provide a bit more analysis. The field of NLP within AI and ML has exploded get a lot more papers all the time. This video can help you understand how AlphaCode works and what some of the key takeaways are.


youtube: https://www.youtube.com/watch?v=t3Yh56efKGI
blog post: https://deepmind.com/blog/article/Competitive-programming-with-AlphaCode
paper: https://storage.googleapis.com/deepmind-media/AlphaCode/competition_level_code_generation_with_alphacode.pdf

🐙 OCTIS : Optimizing and Comparing Topic Models is Simple!

Github: https://github.com/mind-Lab/octis

Paper: https://arxiv.org/abs/2202.07631v1

Dataset: https://paperswithcode.com/dataset/20-newsgroups

@ai_machinelearning_big_data

One Voice Detector to Rule Them All

A brief English article about our VAD got released on The Gradient!

Please follow the link to learn:

- Which values we did pursue;
- Why we decided to create our own VAD;
- Which criteria and metrics we optimized;
- A brief overview of what is available in general;
- How it compares with well-established and similar class solutions;

Links:

- The article https://thegradient.pub/one-voice-detector-to-rule-them-all/
- The VAD is always available on Github (please give us a ⭐️) here - https://github.com/snakers4/silero-vad

PS

- Also new features probably will be reserved for later quarters, but you can vote here
- Also you can find a Russian article here

How Useful is Self-Supervised Pretraining for Visual Tasks?

A relatively old paper (CVPR2020), by our fast life standards. Nevertheless, it has a pair of practical takeaways.

Authors created a synthetic dataset with several degrees of freedom to vary difficulty. It varies from almost monochrome objects to randomized textures and positioning on image.

The target was to compare how good different self-supervised approaches help to tune for different downstream tasks. From classification to depth estimation.

Two practical takeways are:
1. The self-supervised method utility is wildly dependent on task, markup amount and even data complexity.
2. A linear evaluation score, so popular in papers, has almost no correlation with actual fine-tuning results.

Authors found out, that there is no improvement by self-supervised training when lots of labeled data presented (which became kinda well known since then). Based on this, they hypothesise, that improvement of SSL pre-training is rather kind of a regularization than optimization. That is, SSL pre-training helps to find wider optimum, not better. Though, to claim this, some kind of loss plane investigation would be more helpful.

Source: here

Models based on graphs are quite important for a lot of tasks in NLP. There is an overview from Michael Bronstein about what he is expecting for upcoming year for the Graph ML field:

1. Geometry becomes increasingly important in ML.
2. Message passing is still the dominant paradigm in GNNs.
3. Differential equations give rise to new GNN architectures.
4. Old ideas from Signal Processing, Neuroscience, and Physics get a new life.
5. Modeling complex systems requires going beyond graphs.
6. Reasoning, axiomatisation, and generalisation are still big open questions in Graph ML.
7. Graphs become increasingly popular in Reinforcement Learning, but probably still have a way to go.
8. AlphaFold 2 is a triumph of Geometric ML and a paradigm shift in structural biology.
9. Drug discovery and design benefits from GNNs and their confluence with Transformers.
10. AI-first drug discovery is increasingly using Geometric and Graph ML.
11. Quantum ML benefits from graph-based methods.

[link]

Detection of COVID-19 using multimodal data from a wearable device: results from the first TemPredict Study


Some time ago in a different world one of the channel editors shared permmission to use data from sleep & activity tracker Oura Ring to develop an algorithm for COVID-19 prediction.

Results of this study continue to arrive. Today team shared the second manuscript from the first TemPredict Study in Nature Scientific Reports. This manuscript details an algorithm designed to detect COVID-19 using data from the Oura Ring. Alogirthm publication: www.nature.com/articles/s41598-022-07314-0 

The first publication from the first TemPredict Study will continue to be available online for you to access at any time, at this link: https://www.nature.com/articles/s41598-020-78355-6

The first publication from the second TemPredict Study (correlations between data from the Oura Ring and data from a LabCorp antibody blood test) will also continue to be available online for you to access at any time, at this link: https://www.mdpi.com/2076-393X/10/2/264

That's the power of the international collaboration 💪

#oura #covid #biolearning #medical #health

This meme is stolen.

🗿 StyleNeRF: A Style-based 3D-Aware Generator for High-resolution Image Synthesis

Github: https://github.com/facebookresearch/StyleNeRF

Video: https://jiataogu.me/style_nerf

Paper: https://arxiv.org/abs/2110.08985

Project: https://jiataogu.me/style_nerf/

Dataset: https://github.com/facebookresearch/StyleNeRF#dataset

@ai_machinelearning_big_data

Deep Neural Nets: 33 years ago and 33 years from now

Great post by Andrej Karpathy on the progress #CV made in 33 years.

Author's ideas on what would a time traveler from 2055 think about the performance of current networks:

* 2055 neural nets are basically the same as 2022 neural nets on the macro level, except bigger.
* Our datasets and models today look like a joke. Both are somewhere around 10,000,000X larger.
* One can train 2022 state of the art models in ~1 minute by training naively on their personal computing device as a weekend fun project.
* Today’s models are not optimally formulated, and just changing some of the details of the model, loss function, augmentation or the optimizer we can about halve the error.
* Our datasets are too small, and modest gains would come from scaling up the dataset alone.
* Further gains are actually not possible without expanding the computing infrastructure and investing into some R&D on effectively training models on that scale.


Website: https://karpathy.github.io/2022/03/14/lecun1989/
OG Paper link: https://yann.lecun.com/exdb/publis/pdf/lecun-89e.pdf

#karpathy #archeology #cv #nn

💭 NeuralSpeech is a research project in focusing on neural network based speech processing

Github: https://github.com/microsoft/NeuralSpeech

Paper: https://arxiv.org/pdf/2109.14420v3.pdf

Speech Research: https://speechresearch.github.io/

Dataset: https://paperswithcode.com/dataset/aishell-1

@ai_machinelearning_big_data

🦾 Global Tracking Transformers

Github: https://github.com/xingyizhou/GTR

Demo: https://github.com/facebookresearch/detectron2/blob/main/GETTING_STARTED.md

Paper: https://arxiv.org/abs/2203.13250v1

Dataset: https://paperswithcode.com/dataset/mot17

https://t.iss.one/ArtificialIntelligencedl

🔍 Exploiting Explainable Metrics for Augmented SGD

A new explainability metrics that measure the redundant information in a network's layers and exploit this information to augment the Stochastic Gradient Descent

Project

Code: https://github.com/mahdihosseini/rmsgd

Paper: https://arxiv.org/pdf/2203.16723v1.pdf

Dataset: https://paperswithcode.com/dataset/mhist

@ArtificialIntelligencedl

​​Big step after first DALL·E — DALL·E 2

In January 2021, OpenAI introduced DALL·E. One year later, their newest system, DALL·E 2, generates more realistic and accurate images with 4x greater resolution.


The first DALL·E is a transformer model. It receives both the text and the image as a single stream of data containing up to 1280 tokens, and is trained using maximum likelihood to generate all of the tokens, one after another. This training procedure allows DALL·E to not only generate an image from scratch, but also to regenerate any rectangular region of an existing image that extends to the bottom-right corner, in a way that is consistent with the text prompt.

In the second DALL·E they reformated method and now it is CLIP + diffusion model.
CLIP to encode text prior and diffusion model to decode resulting embeding to high resolution image.
So it’s simply GLIDE, but with some tweaks. To generate high resolution images, they train two diffusion upsampler models.

But the results are amazing. Despite that it is cherry picks of course :))

- paper
- blog with images and demos
- video

Silero TTS V3 Finally Released

We have just released a brand new Russian speech synthesis model.

We have made a number of promises we kept:

- Model size reduced 2x;
- New models are 10x faster (!);
- We added flags to control stress;
- Now the models can make proper pauses;
- High quality voice added (and unlimited "random" voices);
- All speakers squeezed into the same model;
- Input length limitations lifted, now models can work with paragraphs of text;
- Pauses, speed and pitch can be controlled via SSML;
- Sampling rates of 8, 24 or 48 kHz are supported;
- Models are much more stable — they do not omit words anymore;

Next steps:

- Release models for the CIS languages, English, some European languages and Hindic languages
- Even further 2-4x speed up
- Updated stress model
- Phonemes support and and built-in voice transfer

Links:

- GitHub - https://github.com/snakers4/silero-models#text-to-speech
- Colab - https://colab.research.google.com/github/snakers4/silero-models/blob/master/examples_tts.ipynb
- Russian article - https://habr.com/ru/post/660565/
- English article - https://habr.com/ru/post/660571/

Imagen — new neural network for picture generation from Google

TLDR: Competitor of DALLE was released.

Imagen — text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding. Imagen builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation. #Google key discovery is that generic large language models (e.g. T5), pretrained on text-only corpora, are surprisingly effective at encoding text for image synthesis: increasing the size of the language model in Imagen boosts both sample fidelity and image-text alignment much more than increasing the size of the image diffusion model.

Website: https://imagen.research.google

#GAN #CV #DL #Dalle

Hi, our friends @mike0sv and @agusch1n just open-sourced MLEM - a tool that helps you deploy your ML models as part of the DVC ecosystem

It’s a Python library + Command line tool.

TLDR:
📦 MLEM can package an ML model into a Docker image or a Python package, and deploy it to Heroku (we made them promise to add SageMaker, K8s and Seldon-core soon :parrot:).

⚙️ MLEM saves all model metadata to a human-readable text file: Python environment, model methods, model input & output data schema and more.

💅 MLEM helps you turn your Git repository into a Model Registry with features like ML model lifecycle management.

Read more in release blogpost: https://dvc.org/blog/MLEM-release
Also, check out the project: https://github.com/iterative/mlem
And the website: https://mlem.ai

Guys are happy to hear your feedback, discuss how this could be helpful for you, how MLEM compares to MLflow, etc.
Ask in the comments!

#mlops #opensource #deployment #dvc

The Cat is on the Mat

Interesting approach to be combined with Ngram embeddings when span boundaries are fuzzy.

I guess can be used downstream with existing sentence parsers.

Such models can be rough and dirty, cheap to train and robust.

- https://explosion.ai/blog/spancat