Welcome to the September edition of the ‚Machine-Learning-Collage' series, where I provide an overview of the different Deep Learning research streams. So what is a ML collage? Simply put, I draft one-slide visual summaries of one of my favourite recent papers. At the end of the month all of the resulting visual collages are collected in a summary blog post. Thereby, I hope to give you a visual and intuitive deep dive into some of the coolest trends.

Sooner or later, the concept of digitization will completely take over all repetitive tasks. Today, with the help of big data, advanced technologies like automation, artificial intelligence, IoT, and machine learning are leveraging unimaginable amounts and types of information to work from. It is streamlining tedious, repetitive, and difficult tasks, which tend to slow down production and also increases the cost of operation. Owing to the evolution of technology, artificial intelligence startups are mushrooming like never before. The companies are driving the world into a new phase of digitization with a mixture of disruptive statistical methods, computational intelligence, soft computing, and traditional symbolic AI. Artificial intelligence is the combination of two amazing concepts namely science and engineering. With the infusion of disruptive trends and human intelligence, intelligent machines and intelligent computing programs are emerging. Slowly, the flare of innovations moved away from IT and entered into diverse industries including healthcare, education, finance, marketing, business, telecommunication, etc. Organizations realized that by digitizing repetitive tasks, an enterprise can cut the cost of paperwork and labor which further eliminates human error, thus boosting efficiency. Automating processes involve employing artificial intelligence solutions that can support digitization and deliver data-driven insights. Artificial intelligence startups emerge as a ready-made solution provider that supports every company's individual needs. AI startups in 2021 use big data to sophisticated AI models and leverage new solutions that could better serve customers. Analytics Insight has listed the top 100 artificial intelligence startups that are driving the next-generation development in technology. It democratizes the way investments are done by bringing sophisticated elite trading technology to laymen. Accrad is a health tech company that assists radiologists to reduce their workload with the precision of artificial intelligence. Radiologists work under different circumstances and deadlines and might find diagnosis through x-rays a bit difficult. Therefore, Accrad has come up with a futuristic solution to help with accurate and fast image diagnosis. The company has made x-ray processing more convincing and simpler. Its signature product CheXRad, a deep learning algorithm that identifies locations in the chest radiograph has the capability to predict 15 different diseases including Covid-19. Affable.ai is a data-driven influencer marketing platform where customers can find relevant and authentic influencers and manage marketing operations. By using cutting-edge computer vision algorithms on social media posts, the company delivers actionable insights about micro-influencers and their audience. Similar to how Google has sophisticated its search and promote relative ads to users, Affable.ai has also built one-click marketing at a shorter scale.

TrackMate is an automated tracking software used to analyze bioimages and distributed as a Fiji plugin. Here we introduce a new version of TrackMate rewritten to improve performance and usability, and integrating several popular machine and deep learning algorithms to improve versatility. We illustrate how these new components can be used to efficiently track objects from brightfield and fluorescence microscopy images across a wide range of bio-imaging experiments.

Since OpenAI released CLIP, trained on internet pictures and their nearby text, people have been using it to generate images. In all these methods - CLIP Dall-E, CLIP BigGAN, CLIP FFT, CLIP VQGAN, CLIP diffusion - you come up with a text prompt, some algorithm presents its images to CLIP, and CLIP's role is to judge how well the images match the prompt. With CLIP's judgements for feedback, the algorithm can self-adjust to make its images match the prompt. But we also do the reverse and set up an app where you give CLIP an image, and then CLIP judges how well text matches the image. One such app is CLIP backpropagation.

As we have previously described, DL Workbench is a tool that allows you to import Deep Learning models, evaluate their performance and accuracy, and perform different optimization tasks, like calibration for 8-bit integer inference. Profiling and model optimization are device-specific, therefore, to achieve maximum performance in a deployment environment, we need to perform these steps directly in that environment. DL Workbench helps you with accessing those capabilities on remote machines. Please note that if you want to have access to numerous hardware configurations ready for work and you do not have them locally or in your private lab, you can run DL Workbench in the Intel DevCloud for the Edge, where you can easily start experiments with available hardware. In this paper, we primarily focus on the case when you prepare the model for deployment and need to benchmark it on a specific hardware setup available in your private lab or a pre-production sandbox.

Deep learning is solving biology's deepest secrets at breathtaking speed. Just a month ago, DeepMind cracked a 50-year-old grand challenge: protein folding. A week later, they produced a totally transformative database of more than 350,000 protein structures, including over 98 percent of known human proteins. Structure is at the heart of biological functions. The data dump, set to explode to 130 million structures by the end of the year, allows scientists to foray into previous "dark matter"--proteins unseen and untested--of the human body's makeup.

Researchers at the Washington University School of Medicine use artificial intelligence (AI) deep learning to classify common brain tumors with a high degree of accuracy using a single magnetic resonance imaging (MRI) scan. The new peer-reviewed study has been accepted for publication in Radiology: Artificial Intelligence. "To the best of our knowledge, this is the first study to address the most common intracranial tumor-types and directly determine the tumor class as well as detect the absence of tumor from a 3D MR volume," wrote the researchers. Last year there were over 308,000 new cases of brain and nervous systems cancer, and more than 250,000 deaths worldwide according to the Global Cancer Statistics (GLOBOCAN) 2020 report. In the United Kingdom, over 11,000 people are diagnosed with a primary brain tumor annually, according to the National Health Service (NHS).

In order to ascend from one rung to the next, you need to develop a completely new set of skills. You can't just make an imitation system larger and expect it to suddenly be able to reason. Yet this is what we are currently doing with our ever-increasing deep learning models: We think that by giving them more power to imitate, they will at some point magically develop the ability to think. Apart from self-delusional hope and selling nice stories to investors and newspapers, there is little reason to believe that.

Transformer-based pre-trained language models boost the performance of open-domain dialogue systems. Prior works leverage Transformer-based pre-trained language models to generate texts with desired attributes in two general approaches: (1) gradient-based methods: updating all latent representations of pre-trained models with gradients from attribute models; (2) weighted-decoding methods: re-ranking beam candidates from pre-trained models with attribute functions. However, gradient-based methods lead to high computation cost and can easily get overfitted on small training sets, while weighted-decoding methods are inherently constrained by the low-variance high-bias pre-trained model. In this work, we propose a novel approach to control the generation of Transformer-based pre-trained language models: the SideControl framework, which leverages a novel control attributes loss to incorporate useful control signals, and is shown to perform well with very limited training samples. We evaluate our proposed method on two benchmark open-domain dialogue datasets, and results show that the SideControl framework has better controllability, higher generation quality and better sample-efficiency than existing gradient-based and weighted-decoding baselines.

Pre-training visual and textual representations from large-scale image-text pairs is becoming a standard approach for many downstream vision-language tasks. The transformer-based models learn inter and intra-modal attention through a list of self-supervised learning tasks. This paper proposes LAViTeR, a novel architecture for visual and textual representation learning. The main module, Visual Textual Alignment (VTA) will be assisted by two auxiliary tasks, GAN-based image synthesis and Image Captioning. We also propose a new evaluation metric measuring the similarity between the learnt visual and textual embedding. The experimental results on two public datasets, CUB and MS-COCO, demonstrate superior visual and textual representation alignment in the joint feature embedding space