IBM Research Editorial Staff

TAPAS: Frugally Predicting the Accuracy of a Neural Network Prior to Training

Constructing a neural network model for each new dataset is the ultimate nightmare for every data scientist. What if you could forecast the accuracy of the neural network earlier thanks to accumulated experience and approximation? This was the goal of a recent project at IBM Research and the result is TAPAS or Train-less Accuracy Predictor […]

Continue reading

AI Year in Review: Highlights of Papers and Predictions from IBM Research AI

For more than seventy years, IBM Research has been inventing, exploring, and imagining the future. We have been pioneering the field of artificial intelligence (AI) since its inception. We were there when the field was launched at the famous 1956 Dartmouth workshop. Just three years later, an IBMer and early computer pioneer, Arthur Samuel, coined […]

Continue reading

Efficient Deep Learning Training on the Cloud with Small Files

Here I describe an approach to efficiently train deep learning models on machine learning cloud platforms (e.g., IBM Watson Machine Learning) when the training dataset consists of a large number of small files (e.g., JPEG format) and is stored in an object store like IBM Cloud Object Storage (COS). As an example, I train a […]

Continue reading

Interpretability and Performance: Can the Same Model Achieve Both?

Interpretability and performance of a system are usually at odds with each other, as many of the best-performing models (viz. deep neural networks) are black box in nature. In our work, improving simple models, we try to bridge this gap by proposing a method to transfer information from a high-performing neural network to another model […]

Continue reading

Delta-Encoder: Synthesizing a Full Set of Samples From One Image

Delta-encoder is a novel approach for few- and one-shot object recognition, in which a modified auto-encoder (called delta-encoder) extracts transferable intra-class deformations (deltas) between same-class pairs of training examples, then applies them to a few examples of a new class (unseen during training) to efficiently synthesize samples from that class. The synthesized samples are then […]

Continue reading

Scalable Forecasts for IoT in the Cloud

This week at the International Conference on Data Mining, IBM Research-Ireland scientist Francesco Fusco demonstrated IBM Research Castor, a system for managing time series data and models at scale and on the cloud. Businesses of today run on forecasts. Whether a hunch of what we think is going to happen or the product of carefully […]

Continue reading

IBM Research AI at 2018 Conference on Neural Information Processing Systems

At the Thirty-Second Conference on Neural Information Processing Systems in Montreal, IBM Research AI will share new ideas and results across our portfolio of research aimed at progressing AI towards real-world challenges. Throughout the week, we will present dozens of papers and demos showcasing our work, as listed below. In addition, we will highlight three […]

Continue reading

Enabling Fabrication Beyond 7nm

How did we get from the Palm Pilots of the 90s to the ultra-powerful smart phones of today? In large part, because of scaling, where integrated circuits are made with smaller feature sizes fitting more and more circuit elements in the same area of silicon at each technology generation. This sets our expectations that in […]

Continue reading

The Pockels Effect on Silicon: A New Material for Ultra-Fast Data Transfer

As data grows, so does the energy consumption required to store and process it, which is why scientists around the world are turning to light (photonics) as a means of moving data. The Pockels effect, an important phenomenon in certain materials, allows changing the optical properties of the material through an electrical stimulus. The effect […]

Continue reading

Addressing Climate Change at a Major Source: Buildings

Can silica gel desiccant packs, often found in shoe boxes and electronics, be the answer to our energy challenges for buildings? A four-year project proves yes, sort of. According to the European Commission, heating and cooling in buildings and industry accounts for half of the EU’s energy consumption, and fossil fuels generate 84 percent of […]

Continue reading

Evolving Speech and AI as the Window into Mental Health

Mental health and neurological disorders are a growing epidemic. In the U.S., nearly one in every five people has a mental health condition (1). Yet there is a growing shortage of mental health professionals to adequately treat this need. By 2025, it’s estimated that demand for psychiatrists may outstrip supply by up to 15,600 psychiatrists […]

Continue reading

Learning Chinese-Specific Encoding for Phonetic Similarity

Performing the mental gymnastics of making the phoenetic distinction between words and phrases such as “I’m hear” to ‘I’m here’ or “I can’t so but tons” to “I can’t sew buttons,” is familiar to anyone who has encountered autocorrected text messages, punny social media posts and the like. Although at first glance it may seem […]

Continue reading