Over the past few years, deep learning has become another trendy word. It is mostly used when the conversation is about machine learning, artificial intelligence, big data, analytics, etc. Currently, it is showing great promise when it comes to developing the autonomous, self-teaching systems that are revolutionizing many industries. Therefore, I decided to write an article about deep learning startups, use cases, and books.
Deep learning was developed as a machine learning approach to deal with complex input-output mappings. Deep learning crunches more data than machine learning — and that is the biggest difference. Basically, if you have a little bit of data, machine learning is a good choice, but if you have a lot of data, deep learning is a better choice for you. Deep learning algorithms do complicated things, like matrix multiplications. They also learn high-level features, so in the case of facial recognition, the algorithm will get the image pretty close to the raw version in replication, whereas machine learning’s images would be blurry. Another powerful feature is that it forms an end-to-end solution instead of breaking a problem and solution down into parts.
What Is Deep Learning?
But what is deep learning exactly? Why has it become so popular? In simple words, deep learning carries out the machine learning process using an artificial neural net that is composed of a number of levels in a hierarchy. For example, the network learns something simple at the initial level in the hierarchy and then sends this information to the next level. The next level takes this simple information, combines it to create something that is a bit more complex, and passes it on the third level. This process continues as each level in the hierarchy builds something more complex from the input it received from the previous level.
Taking an example of a picture of a dog, the initial level of a deep learning network might use differences in the light and dark areas of an image to learn where edges or lines are. The initial level passes this information about edges to the second level, which combines the edges into simple shapes like a diagonal line or a right angle. The third level combines the simple shapes into more complex objects likes ovals or rectangles. The next level might combine the ovals and rectangles into paws and tails. The process continues until it reaches the top level in the hierarchy, where the network has learned to identify dogs. While it was learning about dogs, the network also learned to identify all of the other animals it saw along with the dogs. It is a very good option to identify errors. In general, it is a very fast and efficient way to analyze a huge amount of information and save costs.
Deep Learning Use Cases
Just like we mentioned, deep learning startups successfully apply it to big data for knowledge discovery, knowledge application, and knowledge-based prediction. In other words, deep learning can be a powerful engine for producing actionable results. A good way to see all the potential of deep learning is looking at deep learning startups and see how big companies apply and use it.
Let’s start with the most known examples, deep learning is heavily used by Google in its voice and image recognition algorithms. Also, it is used by Netflix and Amazon to decide what you want to watch or buy next, and by researchers at MIT to predict the future.
How Do Companies Use Deep Learning?
- Automatic speech recognition. Just like we mentioned above, this is one of the most known features of deep learning and big brands use it heavily. For example, Microsoft Cortana, Skype Translator, Amazon Alexa, Google, and Apple Siri, are based on deep learning.
- Image recognition. As people prefer visual stuff, image recognition has gained traction. It is used to analyze documents and pictures connected to a large database, and to make sure that fraud is avoided.
- Natural language processing. Natural language processing is another trendy topic and I even wrote an article about it. It is used by different companies in many industries, especially for negative sampling, word embedding, sentiment analysis, spoken language understanding, machine translation, contextual entity linking, and writing style recognition.
- Drug discovery and toxicology. There are deep learning neural networks for structure-based rational drug design. Researchers enhanced deep learning for drug discovery by combining data from a variety of sources. Now, deep learning is used to predict novel candidate biomolecules for several disease targets, most notably treatments for the Ebola virus.
- Customer relationship management. Deep learning is used a lot in direct marketing for CRM automation. It is good to approximate the value of possible direct marketing actions over the customer state lifetime value.
- Recommendation systems. Recommendation systems use deep learning to extract meaningful features for recommendations. It has been applied for learning user preferences from multiple domains.
- Bioinformatics. It is also used to predict gene ontology annotations, gene-function relationships, and sleep quality based on data from wearables and predictions of health complications from electronic health record data.
- Gesture recognition. Gesture recognition is the latest addition in the area of machine learning that deals with recognizing the gestures made by the human face. The signals emitted from sensors are able to detect the emotion based on energy, time delay, and frequency shifts. It is also able to identify the object and its characteristics.
10 Deep Learning Startups
Deep learning startups come up with absolutely amazing ideas and projects. Let’s look at the brightest ones. These examples are just a small sample of the many companies that are using deep learning to do innovative and exciting things.
1. Bay Labs
Bay Labs is the first one on my list of deep learning startups. It is among the startups applying deep learning to medical imaging to help in the diagnosis and management of heart disease. They want to push the limits of deep learning to make an impact on healthcare. By improving access, value, and quality of medical imaging, they hope to promote and advance healthcare in both the developed and developing world. At Bay Labs, they believe that deep learning has potential to dramatically impact the leading cause of death, cardiovascular disease.
Canary is a New York City-based deep learning startup with a mission to make people safer in and more connected to their homes. Canary is the world’s first smart home security device for everyone. Canary contains an HD video camera and sensors that track everything from temperature and air quality to vibration, sound, and movement. It is controlled entirely from your smartphone. Canary alerts you when it senses anything out of the ordinary, from sudden temperature spikes that can indicate a fire to sound and vibration that could mean an intrusion. Over time, Canary learns your home’s rhythms to send even smarter alerts. Watch a video about Canary here.
3. Knit Health
Knit Health is a sleep vision company whose mission is to help families sleep better and stay healthier. Combining novel computer vision and deep learning technologies, Knit can provide families with personalized insights, suggestions, and risk factors about what happens when they sleep at night, all with just a camera. Knit is currently working on replacing the need for a sleep lab, providing a human-centered and clinically accurate platform for sleep management. Knit’s sleep platform has the ability to learn and track critical markers of sleep issues from breathing to sleep quality to nighttime behaviors, all without wearables or wires. With clinical accuracy, Knit can turn this data into actionable insights for both families and doctors to help in the assessment and treatment of sleep issues.
BenchSci is a machine learning platform that helps biomedical researchers find the best biological compounds for their experiments. It was born as a result of the common struggle with browsing millions of scientific publications in order to find the antibodies best suited for our experiments. BenchSci is a platform that extracts usage evidence from scientific papers and organizes it around antibodies. With BenchSci, scientists can find the best antibodies within minutes. Watch a video about BenchSci here.
CarePredict is designed to solve a specific challenge in senior care: family, friends, and caregivers of an elderly person may not notice the precursors to declines in health and hence do not intervene in time, leading to hospital admissions of easily preventable issues. For example, a senior entering into a depressive phase will start having restless sleep patterns, loss of hygiene, and changes in eating patterns several days before the episode. CarePredict solves the continuous observation problem for the senior market with the very first wearable designed for seniors that tracks their activities of daily living — from waking up, bathing, sleeping, and quality of sleep to brushing teeth, eating, drinking, cooking, and more. It gives useful insights. Watch a video about CarePredict here.
GrokStyle is a deep learning AI company. GrokStyle is developing software for visual searches to enable instance recognition of an object. Their mission is to bridge the gap between inspiration and retail. They are experts in search and recommendation. The techniques they are developing can be applied broadly to domains like interior design, apparel search, real estate search, product lookup, etc. Given a photo, they answer questions like “What is this product?” and “Where can I buy it?” and “What goes with this?” GrokStyle was recently named one of the top 100 most promising private AI companies globally by CB Insights. Watch a video about GrokStyle here.
Drive.ai is a Silicon Valley deep learning startup founded by former lab mates out of Stanford University’s Artificial Intelligence Lab. They are creating Deep Learning for Autonomous Vehicles. They started this project because they believe that this technology has the potential to save lives and transform industries. Watch a video about it here.
Enway develops the software stack for autonomous service vehicles. They believe in teaming up autonomous vehicles and human labor to make jobs like street sweeping or trash collection safer, easier, and more efficient.
ViSenze simplifies search and categorization in your image database with visual search and image recognition via an API integration. It develops commercial applications that use deep learning networks to power image recognition and tagging. Customers can use pictures rather than keywords to search a company’s products for matching or similar items. Media owners and brands use ViSenze to turn images into immediate engagement opportunities such as product recommendations and Ad targeting.
And the last one on the list of promising deep learning startups is Atomwise. It applies deep learning networks to the problem of drug discovery. Atomwise uses deep learning networks to help discover new medicines, to explore the possibility of repurposing known and tested drugs for use against new diseases.
Deep Learning Books That Are Worth Reading
Lastly, let's look over a few deep learning books you might want to read.
"Deep Learning" by Ian Goodfellow, Yoshua Bengio, and Aaron Courville
Deep learning is a form of machine learning that enables computers to learn from experience and understand the world in terms of a hierarchy of concepts. This book offers mathematical and conceptual background, covering relevant concepts in linear algebra, probability theory, information theory, numerical computation, and machine learning. It describes deep learning techniques used by practitioners in industry, including deep feedforward networks, regularization, optimization algorithms, convolutional networks, sequence modeling, and practical methodology; and it surveys such applications as natural language processing, speech recognition, computer vision, online recommendation systems, bioinformatics, and video games. Also, the book offers research perspectives, covering such theoretical topics as linear factor models, autoencoders, representation learning, structured probabilistic models, Monte Carlo methods, the partition function, approximate inference, and deep generative models.
"Deep Learning: A Practitioner’s Approach" by Josh Patterson and Adam Gibson
Reading this book, you will dive into machine learning concepts in general, as well as deep learning in particular. You will understand how deep networks evolved from neural network fundamentals and you will explore the major deep network architectures, including convolutional and recurrent. Also, you will learn how to map specific deep networks to the right problem.
"Fundamentals of Deep Learning: Designing Next-Generation Machine Intelligence Algorithms" by Nikhil Buduma and Nicholas Locascio
Deep learning has become an extremely active area of research. In this practical book, authors provide examples and clear explanations to guide you through major concepts of this complicated field. Companies such as Google, Microsoft, and Facebook are actively growing in-house deep learning teams. However, deep learning is still a pretty complex and difficult subject to understand. This book will give you a solid foundation of deep learning understanding.