Day 1 (21.11.2017):
The FDP started with an inauguration by Dr.Angelina Geetha-Professor, Crescent University. She shared her academic experience in a wide variety of domains. Dr. Krishnamurthy, HOD of CSE delivered welcome address and Principal Dr.G.Prabhakaran gave presidential address and session was started for the day.
Session 1 : Deep Learning and Big Data Analytics
Speaker: Dr.Angelina Geetha, Professor – CSE Crescent University
Dr.Angelina Geetha gave a detailed introduction to what is data, information, knowledge and understanding. She also explained how analytics is different from analysis. She briefly introduced what Wisdom of things is and introduced data mining. She explained different machine learning methods like classification, clustering, association rule discovery, sequential rule discovery, regression and deviation detection. Her insight on Hadoop framework and Map reduce technique was very useful. She also explained the 4 types of analytics – descriptive, diagnostic, predictive and prescriptive. Her real time applications and examples made the session very interesting and concepts easy to understand. She introduced deep learning and how neural networks help in achieving deep learning. She explained multi-layer perceptron, training set data and testing set data in detail. She gave a quick overview of an open source tool called H2O for deep learning. The first session gave an awesome start for the 2 day workshop.
Session 2 : Machine Learning Algorithms and Commercial Applications.
Speakers: Mr Madhan U Srinivasan & Raghavan Subramanian, TechArtus Solutions
Mr.Madhan gave an insight on the difference between with programming and machine learning. He explained classification in detail with examples. Explained how neural networks. He explained the machine learning need in the healthcare domain and the practical challenges in developing solutions in this domain. He showed lot of videos to emphasis of machine learning for the current scenario. He opened up a lot of areas where we could start research work on. He gave an quick introduction to Tessarach a HP tool and other open source tool for machine learning. The second part of the session was handled by Mr.Raghavan, founder of TechArtus Solutions, Chennai. He gave a detail insight on Simiyin their proprietary tool for machine learning. A Live demo regarding prediction of stock market was very useful.
Day 2(22.11.2017):
Session 1: Visualisation in Deep Learning
Speaker: Mr.Dhanasekar Paneerselva-Senior Lead, HCL technologies
Mr.Dhanasekar Paneerselvam was the keynote speaker on the Topic. He gave a detailed insight on Business Analytics, Data Analytics and its types. He explained the Analytic process, Analytical examples and challenges. He also gave a detailed insight on the topic through a case study on Google Analytics. He guided us with the tools that will help in data analytics. Then he focused on what is visualization and what is visual analytics. He explained visual analytics model and detailed the importance of data visualization. This session was indeed very useful and informative on how data analytics is used practically used in so many areas.
Session 2: Deep Learning in Neural Networks
Speaker: Dr.Krishnamurthy, HOD of CSE department, KCG College of Technology
Dr.Krishnamurthy gave introduction on Neural Network and how multilayered architecture helps in deep learning. He had a very interactive and interesting talk about single,2 and 3 tier architecture and explained in detail how multilayer architecture helps in refining data. He also explained about feed forward and back propagation. He explained how the hidden layers satisfy the unsupervised learning of deep learning. He explained the difference and significance of training data and test data. He also gave a example problem to understand how weights of the edges influence in the output calculation. Overall this was the most interactive session.
Session 3: Deep Learning with Tensorflow
Speaker: Mr.Robert Jean, Senior Embedded Engineer, Vistronics Design Solutions
Mr.Robert Jean gave quick technical theory introduction to deep learning through the usual Cat/Dog Classifier. He explained about back propagation and how training data is important for better results. Tensorflow is an ope source tool developed by Google to write our own deep learning algorithms. Mr.Robert helped us install the tensor flow so that we can have an hands on. But since we did not have the right installable, we could do it. So he has a live demo of a sample program in python to implement a deep learning algorithm. Audiences were excited to get his guidance on how to write, execute and visualize the graph.
Session 4: Domain Specific Applications of Visual Analysis of Deep Neural Networks.
Speaker: Dr.S.Sathyalakshmi, Professor of CSE, HITS
Dr.S.Sathyalakshmi gave different areas where deep learning has a great future. She gave details on different areas like visual sentiment analysis, Colorization of black and white photo, Automatic machine translation, Object classification and detection, Automatic Handwriting Generation, Automatic Image Caption Generation and Automatic text generation. This session was useful as it gave an insight so that we can chose our areas of interest and do our future research.