Neural language models are becoming the prevailing methodology for the tasks of query answering, text classification, disambiguation, completion and translation. Commonly comprised of hundreds of millions of parameters, these neural network models offer state-of-the-art performance at the cost of interpretability; humans are no longer capable of tracing and understanding how decisions are being made. The attention mechanism, introduced initially for the task of translation, has been successfully adopted for other language-related tasks. We propose AttViz, an online toolkit for exploration of self-attention---real values associated with individual text tokens. We show how existing deep learning pipelines can produce outputs suitable for AttViz, offering novel visualizations of the attention heads and their aggregations with minimal effort, online. We show on examples of news segments how the proposed system can be used to inspect and potentially better understand what a model has learned (or emphasized).
Data science is a very wide field, and one of the promising fields that is spreading in a fast way, also, it is one of the very rewarding, and it is increasing in expansion day by day, due to its great importance and benefits, as it is the future. Companies can analyze trends to make critical decisions to engage customers better, enhance company performance, and increase profitability. And the employment of data science and its tools depends on the purpose you want from them. For example, using data science in health care is very different from using data science in finance and accounting, and so on. And I'll show you the core libraries for data handling, analysis and visualization which you can use in different areas.
We formulate data visualization as a sequence to sequence translation problem. TLDR; We train a model that can take in a dataset as input and generate a plausible visualizations as output. This work is done jointly with a colleague (Cagatay Demiralp) and started from a conversation we had after a paper discussion meeting. We had read some papers where various forms of generative models were used to create a wide range of stuff -- from generating images (GANs), music, source code etc to generating questions and answers about images (VQA) etc. Despite the quirks that can sometimes be associated with generative models (one eyed cats, music that ultimately lacks that natural feel etc), they all demonstrate a promise of value when trained and deployed at scale.
At first glance, data science always appears to be an intricate field -- or maybe I should say a collection of fields. It very broad vague, and one can argue complex. But, the truth is, data science can be defined very simply using one sentence. Data science is the field of interpreting data collected from different resources into useful information. Or in other words, it is all about listening and translating the story some data is trying to deliver.
QlikView developers (as per the book QlikView 11 for Developers) were those of us who wrote load scripts, designed data models, formulated expressions, and manipulated QlikView objects. Qlik Help has now left that group of people nameless and deemed developers to be those who work with either QlikView or Qlik Sense APIs using some third-party code. However, Qlik Sense APIs are at the forefront of what the software is and the title of Qlik Sense developer implies some ability to work with them. Can QlikView developers upgrade their skills and become full-fledged Qlik Sense developers? After some reflection on my days as a QlikView developer and some cheerleading to motivate myself to make this transition, I'm going to share with you my plans to learn the answer.