Deploy Deep Learning Models with Flask

Introduction Nowadays it is easy to build - train and tune - powerful machine learning (ML) models using tools like Spark, Conda, Keras, R etc. The business value of these models, however, only comes from deploying the models into production. Deploying Machine Learning models in production is still a significant challenge. There is no general strategy that fits every ML problem and/or every company’s need. Deployment can be done in wide variety of ways, which entails either loading the model directly into the application or making API’s and calling them from the application.

Read more

Share

Generative Adversarial Networks

\(\color{red}{\text{ Draft: Work in Progress..... }}\) GAN consist of two competing models striving to outdo each other: the Generator and Discriminator models. The Generator takes in random input and tries to generate real data (curves, images, texts, ). The Discriminator is a binary classifier. It takes, as its input, the fake data generated by the Generator, and the real dataset and learn to tell whether the data is real or fake.

Read more

Share

Time Series Forecasting using LSTM in R

Brief Introduction Load the neccessary libraries & the dataset Data preparation Modeling In mid 2017, R launched package Keras, a comprehensive library which runs on top of Tensorflow, with both CPU and GPU capabilities. I highlighted its implementation here. In this blog I will demonstrate how we can implement time series forecasting using LSTM in R. Brief Introduction Time series involves data collected sequentially in time. I denote univariate data by $x_{t} \in \mathbb{R} $ where $t \in \mathcal{T} $ is the time indexing when the data was observed.

Read more

Share

Overview of Machine Learning Algorithms

\(\color{red}{\text{ Draft: Work in Progress..... }}\) Learning Styles: Supervised Learning The values of target variable, in training set are known. Unsupervised Learning data is not labeled and outcomes are unknown. Seek to find hidden structure in unlabeled data Example: Clustering algorithms Semi-supervised Learning Uses both labeled and unlabeled data to improve supervised learning. Example: in Deep learning Generative models Problems: Classification problem: The target variable is qualitative (categorical/nominal) Binary- Two possible outcomes only e.

Read more

Share