Note: Please note that for this project I have worked on notebooks mostly because it involved a lot of exploration and experimentation with various approaches. As such, work is not polished since I am still exploring better ways. I have tried to explain the cells as much as I could. Feel free to reach out to me if something is unclear. I have used Facebook datasets from here. The downloaded graph data is in "mtx" format. This is just another format for sharing matrix data.

A Modern Approach, 3e offers the most comprehensive, up-to-date introduction to the theory and practice of artificial intelligence. Number one in its field, this textbook is ideal for one or two-semester, undergraduate or graduate-level courses in Artificial Intelligence. In this mind-expanding book, scientific pioneer Marvin Minsky continues his groundbreaking research, offering a fascinating new model for how our minds work. He argues persuasively that emotions, intuitions, and feelings are not distinct things, but different ways of thinking. Introduction to Artificial Intelligence presents an introduction to the science of reasoning processes in computers, and the research approaches and results of the past two decades.

In my previous two posts in this series, I've essentially argued both sides of the same issue. In the first, I explained why deep learning is not a panacea, when machine learning systems (now and likely always) will fail, and why deep learning in its current state is not immune to these failures. In the second post, I explained why deep learning, from the perspective of machine learning scientists and engineers, is an important advance: Rather than a learning algorithm, deep learning gives us a flexible, extensible framework for specifying machine learning algorithms. Many of the algorithms so far expressed in that framework give orders of magnitude-level improvement on the performance of previous solutions. In addition, it's a tool that allows us to tackle some problems heretofore unsolvable directly by machine learning methods.

Matrix Factorization (MF) (e.g., Probabilistic Matrix Factorization and NonNegative Matrix Factorization) techniques have become the crux of many real-world scenarios, including graph representation and recommendation system (RecSys) because they are powerful models to find the hidden properties behind the data. The idea behind matrix factorization is to represent users and items in a lower-dimensional latent space. And is widely used in the recommendation system and dimensionality reduction. Although, there are many Python libraries that could perform matrix factorization, building the algorithm from scratch could be helpful to understand the basics. Also, there are many complex cases when the matrix factorization library could not handle.

Again, this is one of the techniques (simplest one) for developing an app using the Django REST Framework on Windows system. This code will help create a Django-project named "video_emotion" with some auto-generated setting files for an instance of Django, including database configuration, Django-specific options, and other application-specific files as shown below. The above command will powerup the machine learning behind our API and creates multiple files inside our app "video_app" folder as shown below. Several of these files aren't even needed and hence, we can ignore those files. Django has a utility that will help us focus on coding than arranging directories for deployment .

AI is not something new. Over the years, it has made immense advancement in every field like healthcare, manufacturing, law, finance, retail, real estate, accountancy, digital marketing, and several other areas. Each one is computational and irrefutable from upcoming changes in the system. AI algorithms have proved dangerous in terms of Skynet images, the matrix, robot Apocalypse, and technological unemployment. A wide range of diverse AI patterns like autonomous systems, chatbots, document classification, advanced predictive analytics solutions have made human labor jobless.

I've lost track of the number of times I've heard someone say Timnit Gebru is saving the world recently. Her co-lead of AI ethics at Google, Margaret Mitchell, said it a few days ago when Gebru led events around race at Google. Gebru's work with Joy Buolamwini demonstrating race and gender bias in facial recognition is one of the reasons lawmakers in Congress want to prohibit federal government use of the technology. That landmark work also played a major role in Amazon, IBM, and Microsoft agreeing to halt or end facial recognition sales to police. Earlier this week, organizers of the Computer Vision and Pattern Recognition (CVPR) conference, one of the biggest AI research events in the world, took the unusual step of calling Gebru's CVPR tutorial illustrating how bias in AI goes far beyond data "required viewing for us all."

Specifying a governing physical model in the presence of missing physics and recovering its parameters are two intertwined and fundamental problems in science. Modern machine learning allows one to circumvent these, via emulators and surrogates, but in doing so disregards prior knowledge and physical laws that are especially important for small data regimes, interpretability, and decision making. In this work we fold the mechanistic model into a flexible data-driven surrogate to arrive at a physically structured decoder network. This provides accelerated inference for the Bayesian inverse problem, and can act as a drop-in regulariser that encodes a-priori physical information. We employ the variational form of the PDE problem and introduce stochastic local approximations as a form of model based data augmentation. We demonstrate both the accuracy and increased computational efficiency of the framework on real world settings and structured spatial processes.

We design a physics-aware auto-encoder to specifically reduce the dimensionality of solutions arising from convection-dominated nonlinear physical systems. Although existing nonlinear manifold learning methods seem to be compelling tools to reduce the dimensionality of data characterized by a large Kolmogorov n-width, they typically lack a straightforward mapping from the latent space to the high-dimensional physical space. Moreover, the realized latent variables are often hard to interpret. Therefore, many of these methods are often dismissed in the reduced order modeling of dynamical systems governed by the partial differential equations (PDEs). Accordingly, we propose an auto-encoder type nonlinear dimensionality reduction algorithm. The unsupervised learning problem trains a diffeomorphic spatio-temporal grid, that registers the output sequence of the PDEs on a non-uniform parameter/time-varying grid, such that the Kolmogorov n-width of the mapped data on the learned grid is minimized. We demonstrate the efficacy and interpretability of our approach to separate convection/advection from diffusion/scaling on various manufactured and physical systems.

Despite the recent success of reinforcement learning in various domains, these approaches remain, for the most part, deterringly sensitive to hyper-parameters and are often riddled with essential engineering feats allowing their success. We consider the case of off-policy generative adversarial imitation learning, and perform an in-depth review, qualitative and quantitative, of the method. Crucially, we show that forcing the learned reward function to be local Lipschitz-continuous is a sine qua non condition for the method to perform well. We then study the effects of this necessary condition and provide several theoretical results involving the local Lipschitzness of the state-value function. Finally, we propose a novel reward-modulation technique inspired from a new interpretation of gradient-penalty regularization in reinforcement learning. Besides being extremely easy to implement and bringing little to no overhead, we show that our method provides improvements in several continuous control environments of the MuJoCo suite.