Artificial Intelligence and Machine Learning

We can provide cutting-edge Artificial Intelligence & Machine Learning capabilities to bespoke Android and web apps!

Whilst Artificial Intelligence and Machine Learning have been around for a while, it is only recently that Deep Learning architectures, such as deep neural networks, are within reach of specialist app developers. Deep Learning architectures allow app developers to integrate Machine Learning (i.e. allowing apps to use data and then learn for themselves) into low-cost bespoke apps. Think of Machine Learning as teaching applications to think like humans whilst being far better than humans with regard to speed of computation and eliminating human error.

Visual Recognition
We have been testing the use of artificial intelligence and machine learning for our bespoke apps and we are simply blown away by what we can achieve in the sphere of image recognition by training deep convolutional neural networks with machine learning algorithms. Google’s Artificial Intelligence unit has made available powerful Machine Learning frameworks like TensorFlow (an open source software library for numerical computation) which are causing a major paradigm shift in the way in which we can give "intelligence" to the bespoke enterprise apps we develop for UK businesses.
Try it!

Voice Recognition
Recognising and distinguishing between different voices within an audio stream when more than one person is talking is very difficult for artificial intelligence (AI). However, Google has developed a new technique of "Speaking Diarization" which is able to achieve circa 92 per cent precision. At the end of October 2018 the basic algorithms were made publicly available in open source form so that they can also be used by third-party developers. We of course took advantage of Google’s largesse and have downloaded what is called their “Unbounded Interleaved-State Recurrent Neural Network (UIS-RNN)” algorithm and are currently experimenting with it. The potential of this technology is enormous, e.g. the AI being able to follow a doctor-patient conversation.
Learn more

Android Neural Networks API
Released in November 2017, the Android Neural Networks API (NNAPI) is used to perform demanding, in-depth, data-processing activities for Machine Learning on smartphones and tablets and contributes a foundation for powerful Machine Learning frameworks like TensorFlow Lite which instruct and teach neural networks. The API is available on smartphones and tablets running Android 8.1 (API level 27) or higher.

NNAPI aids "inferencing" (i.e. deduction, the steps needed to move from assumptions to conclusions) by applying data from Android devices to previously worked out models such as categorising images, forecasting user behavior and providing pertinent responses to a search query.
Learn more

Google Cloud Machine Learning Engine
Google Cloud Machine Learning Engine allows us to create machine learning models that work on all types of data with no size restrictions. It can take any TensorFlow model and perform large scale training (i.e. the software learns from the data) and support thousands of users and many terabytes of data (a terabyte = 1,024 gigabytes).

The service is integrated with Google Cloud Dataflow for pre-processing, allowing us to access data from Google Cloud Storage and Google BigQuery should we wish to. Basically it allows us to harness the power of Google's global data centres and networks to perform high-end Machine Learning which was previously only available to the likes of massive research organisations such as CERN (European Organization for Nuclear Research) when analysing data from the Large Hadron Collider to confirm the existence of the Higgs boson.
Learn more

Some Innovative Uses for Machine Learning & Artificial Intelligence

Assisting Pathologists in Detecting Cancer with Deep Learning
Investigating how deep learning can be applied to digital pathology, by creating an automated detection algorithm that can naturally complement pathologists’ workflow.

Understanding Medical Conversations
Could the voice recognition technologies already available in Google Assistant, Google Home, and Google Translate be used to document patient-doctor conversations and help doctors and scribes summarize notes more quickly?

DeepVariant: Highly Accurate Genomes With Deep Neural Networks
The open source release of DeepVariant, a deep learning technology to reconstruct the true genome sequence from HTS sequencer data with significantly greater accuracy than previous classical methods.

Closing the Simulation-to-Reality Gap for Deep Robotic Learning
Simulating many years of robotic interaction is quite feasible with modern parallel computing, physics simulation, and rendering technology. Moreover, the resulting data comes with automatically-generated annotations, which is particularly important for tasks where success is hard to infer automatically.

A couple of high profile uses of Machine Learning and Tensorflow can be seen below.

Machine Learning algorithm learns chess in 4 hours!
An algorithm developed by Google’s Artificial Intelligence unit, DeepMind, was able to beat world-leading specialist chess software Stockfish 8 just four hours after being given the rules of chess and being told to learn by playing simulations against itself. "Starting from random play, and given no domain knowledge except the game rules, AlphaZero achieved within 24 hours a superhuman level of play in the games of chess and shogi (Japanese chess) as well as Go, and convincingly defeated a world-champion program in each case." Read the paper published by Cornell University on 5 December 2017 for more information.
Learn More

Machine Learning with NASA's Kepler data finds solar system similar to ours!
NASA’s Kepler Space Telescope observed hundreds of thousands of stars between 2009 and 2013 with the aim of identifying planets of a similar size to Earth which were orbiting sun-like stars. Such planets could possibly sustain life and therefore be populated by alien species similar to our own. Engineers at Google Brain trained a deep convolutional neural network using data from the Autovetter Planet Candidate Catalog at NASA's Exoplanet Archive. They then fed the Kepler Space Telescope non-transiting planets data into TensorFlow (an open source software library for numerical computation for machine learning). This then identified two new planets with a high degree of confidence. One planet adding to the five already known planets orbiting Kepler-80 and the other planet adding to the seven already known planets orbiting Kepler-90 making Kepler-90 similar to our solar system (i.e. a sun orbited by 8 planets). The combination of the Harvard College Observatory (HCO) and the Smithsonian Astrophysical Observatory (SAO) is known as the Harvard-Smithsonian Center for Astrophysics (CfA) and they published a very interesting paper on this discovery in December 2017. We have included a link to the actual paper on Harvard-Smithsonian's server.
Learn More

Machine Learning News

Please find below links to recent articles published by Google's "Distill: Supporting Clarity in Machine Learning" initiative which is a new open science journal and ecosystem supporting human understanding of machine learning.

Understanding Convolutions on Graphs
Understanding the building blocks and design choices of graph neural networks.

A Gentle Introduction to Graph Neural Networks
What components are needed for building learning algorithms that leverage the structure and properties of graphs?

Distill Hiatus
After five years, Distill will be taking a break.

Adversarial Reprogramming of Neural Cellular Automata
Reprogramming Neural CA to exhibit novel behaviour, using adversarial attacks.

Weight Banding
Weights in the final layer of common visual models appear as horizontal bands. We investigate how and why.

Branch Specialization
When a neural network layer is divided into multiple branches, neurons self-organize into coherent groupings.

Multimodal Neurons in Artificial Neural Networks
We report the existence of multimodal neurons in artificial neural networks, similar to those found in the human brain.

Self-Organising Textures
Neural Cellular Automata learn to generate textures, exhibiting surprising properties.

Visualizing Weights
We present techniques for visualizing, contextualizing, and understanding neural network weights.

Curve Circuits
Reverse engineering the curve detection algorithm from InceptionV1 and reimplementing it from scratch.

High-Low Frequency Detectors
A family of early-vision neurons reacting to directional transitions from high to low spatial frequency.

Naturally Occurring Equivariance in Neural Networks
Neural networks naturally learn many transformed copies of the same feature, connected by symmetric weights.

Understanding RL Vision
With diverse environments, we can analyze, diagnose and edit deep reinforcement learning models using attribution.

Communicating with Interactive Articles
Examining the design of interactive articles by synthesizing theory from disciplines such as education, journalism, and visualization.

Thread: Differentiable Self-organizing Systems
A collection of articles and comments with the goal of understanding how to design robust and general purpose self-organizing systems.

Self-classifying MNIST Digits
Training an end-to-end differentiable, self-organising cellular automata for classifying MNIST digits.

Curve Detectors
Part one of a three part deep dive into the curve neuron family.

Exploring Bayesian Optimization
How to tune hyperparameters for your machine learning model using Bayesian optimization.

An Overview of Early Vision in InceptionV1
An overview of all the neurons in the first five layers of InceptionV1, organized into a taxonomy of 'neuron groups.'

Visualizing Neural Networks with the Grand Tour
By focusing on linear dimensionality reduction, we show how to visualize many dynamic phenomena in neural networks.

Thread: Circuits
What can we learn if we invest heavily in reverse engineering a single neural network?

Zoom In: An Introduction to Circuits
By studying the connections between neurons, we can find meaningful algorithms in the weights of neural networks.

Growing Neural Cellular Automata
Training an end-to-end differentiable, self-organising cellular automata model of morphogenesis, able to both grow and regenerate specific patterns.

Visualizing the Impact of Feature Attribution Baselines
Exploring the baseline input hyperparameter, and how it impacts interpretations of neural network behavior.

Computing Receptive Fields of Convolutional Neural Networks
Detailed derivations and open-source code to analyze the receptive fields of convnets.

The Paths Perspective on Value Learning
A closer look at how Temporal Difference Learning merges paths of experience for greater statistical efficiency

A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features'
Six comments from the community and responses from the original authors

A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features': Adversarial Example Researchers Need to Expand What is Meant by 'Robustness'
The main hypothesis in Ilyas et al. (2019) happens to be a special case of a more general principle that is commonly accepted in the robustness to distributional shift literature

A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features': Robust Feature Leakage
An example project using webpack and svelte-loader and ejs to inline SVGs

A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features': Two Examples of Useful, Non-Robust Features
An example project using webpack and svelte-loader and ejs to inline SVGs

A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features': Adversarially Robust Neural Style Transfer
An experiment showing adversarial robustness makes neural style transfer work on a non-VGG architecture

A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features': Adversarial Examples are Just Bugs, Too
Refining the source of adversarial examples

A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features': Learning from Incorrectly Labeled Data
Section 3.2 of Ilyas et al. (2019) shows that training a model on only adversarial errors leads to non-trivial generalization on the original test set. We show that these experiments are a specific case of learning from errors.

A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features': Discussion and Author Responses

Open Questions about Generative Adversarial Networks
What we'd like to find out about GANs that we don't know yet.

A Visual Exploration of Gaussian Processes
How to turn a collection of small building blocks into a versatile tool for solving regression problems.

Visualizing memorization in RNNs
Inspecting gradient magnitudes in context can be a powerful tool to see when recurrent units use short-term or long-term contextual understanding.

Activation Atlas
By using feature inversion to visualize millions of activations from an image classification network, we create an explorable activation atlas of features the network has learned and what concepts it typically represents.

AI Safety Needs Social Scientists
If we want to train AI to do what humans want, we need to study humans.

Distill Update 2018
An Update from the Editorial Team

Differentiable Image Parameterizations
A powerful, under-explored tool for neural network visualizations and art.

Feature-wise transformations
A simple and surprisingly effective family of conditioning mechanisms.

The Building Blocks of Interpretability
Interpretability techniques are normally studied in isolation. We explore the powerful interfaces that arise when you combine them -- and the rich structure of this combinatorial space.

Using Artificial Intelligence to Augment Human Intelligence
By creating user interfaces which let us work with the representations inside machine learning models, we can give people new tools for reasoning.

Sequence Modeling with CTC
A visual guide to Connectionist Temporal Classification, an algorithm used to train deep neural networks in speech recognition, handwriting recognition and other sequence problems.

Feature Visualization
How neural networks build up their understanding of images

Why Momentum Really Works
We often think of optimization with momentum as a ball rolling down a hill. This isn't wrong, but there is much more to the story.

Research Debt
Science is a human activity. When we fail to distill and explain research, we accumulate a kind of debt...

Experiments in Handwriting with a Neural Network
Several interactive visualizations of a generative model of handwriting. Some are fun, some are serious.

Deconvolution and Checkerboard Artifacts
When we look very closely at images generated by neural networks, we often see a strange checkerboard pattern of artifacts.

How to Use t-SNE Effectively
Although extremely useful for visualizing high-dimensional data, t-SNE plots can sometimes be mysterious or misleading.

Attention and Augmented Recurrent Neural Networks
A visual overview of neural attention, and the powerful extensions of neural networks being built on top of it.

New Media Aid has been the app development partner for Kelly Turkeys since 2006 and has developed a number of mission-critical bespoke apps for us and we highly recommend them for bespoke application development.

Asa Howard, Finance Director
Kelly Turkeys - Danbury, Essex

New Media Aid has been our development partner for both our website and virtual learning environment since 2004. The quality of service has been consistently exceptional in every way. I would be very happy to recommend them for any development and as a highly reliable long term partner as well as for short term projects.

John Rubinstein, Principal
Woodhouse College - Finchley, London

New Media Aid has developed a number of advanced, bespoke apps for BAE Systems and we are happy to recommend them for bespoke app development.

Rob Morris, Vice President Marketing & Analysis, Asset Management
BAE Systems - Hatfield, Hertfordshire

These apps developed by New Media Aid mark a new turn towards technology for an industry that’s been slow to adapt until now. With no suitable ‘off the shelf’ solution available below costly, enterprise level software, we were involved in the development process from the start, getting the right solution that suited our operations and our customers.

Ben Guy, Managing Director
Hammond Produce

New Media Aid quickly understood what we needed and were able to develop exactly what was required with minimal input from ourselves. As this was the first Android app we had developed we were pleased that it was done on time and within the quoted price. We would gladly work with them again and recommend them.

Mark Austin, Managing Director
Wing Parking

Having recently moved to Android PDAs for our mobile workforce, we identified a need for a Bespoke Android App and a Web-based Management System to control it. New Media Aid demonstrated a real understanding for our needs and a "can do" attitude that we found extremely helpful. Their solution was provided on time and within budget and has exceeded our expectations. We will be happy to work with them again when we have further requirements.

Trevor Wright, IT Manager

New Media Aid has successfully developed a number of bespoke applications for us over the last few years - we have always been pleased with the results and with the ongoing support and assistance we receive.

Anne Barlow, Marketing Manager
British Chiropractic Association - Reading, Berkshire

New Media Aid undoubtedly fulfils the requirements we have for our company’s corporate website. Not only did they provide us with all the features we needed from the get-go, they also promptly come up with solutions when we have new requests.

JD Venter, Deputy Editor
Dry Cargo International - Colchester, Essex

New Media Aid developed a bespoke website and content management system for us that makes it easy to run our business. The app has been designed in such a way that we’re able to have it updated when our needs as a company develop. Our trade customers love the fact that they can login and view historical orders as well as viewing new products and ordering online themselves. I would definitely recommend New Media Aid for bespoke app development.

Shona Easton, Managing Director
Easton Design Studio - Cranleigh, Surrey

We are thrilled and very excited by our new mobile-friendly website from New Media Aid which is seamlessly integrated with a bespoke candidate and vacancy management system. They have been extremely responsive and patient and have talked us through the whole process. They are always available for queries and trouble-shooting not only on this project but also with other web apps they have developed for us. We couldn’t recommend them more highly.

Millie Harvey, Marketing Manager
Excelerate Resources - Warwick

New Media Aid develops business-critical bespoke applications for Gateway College and we are happy to recommend them for bespoke application development.

Paul Taylor, Network And Corporate Data Manager
Gateway College - Leicester

We are delighted with the flexible and efficient service we received from New Media Aid. Highly professional and personable which works well for us as a small business. We wanted to update our old website and not being IT specialists it was good to talk through issues and get helpful advice on new design ideas and practical solutions. Working together on the design and implementation of our new site has resulted in a user-friendly eCommerce website on PCs, phones and tablets – and we instantly saw a huge increase in business once the new site went live. No hesitation in recommending their services.

Gordon Little, Managing Director
GE Little Radiator Covers - Basildon, Essex

New Media Aid developed Gigacom Tems a bespoke and fully integrated project management system. New Media Aid fully support all our IT requirements and must be congratulated on the valued and excellent support that we continue to receive from the New Media Aid team. I have no doubt that without the high level of innovative support that we receive we would not be able to efficiently deliver the services that we provide for our customers.

Phil Taylor, Operational Management Team
Gigacom Telecommunication Engineers Management Services - Alton, Hampshire

The certificate scanning system developed by New Media Aid is working really well for us and is more efficient than our previous way of processing certificates. It really is brilliant and must have saved us a fortune in paper!

Rachel Bothamley, Senior Examinations Officer
Lincoln College

New Media Aid has supported the development of a complex store and back office system that significantly helps improve the accuracy of our processes whilst improving the efficiency and service to our clients. Couldn't be happier with a supplier (who we genuinely consider as a partner).

Stephen Mason, Managing Director
Mason Information Technologies - Nottingham

New Media Aid has an excellent understanding of our business and therefore can offer a high level, intuitive service. They have programmed and maintained our websites and CMS for nearly 10 years and are responsive and very easy to work with, we highly recommend talking to them about your digital projects.

Donald Campbell, Marketing Manager
Team London Bridge - Southwark, London

Excellent service, clear pricing and Alastair is so easy to deal with. Would definitely recommend. We're very pleased with our new App!

Mary Stealey, Operations Manager
Just Filters - Harlow, Essex

Alastair was great at understanding our needs and building a new android app and web-facing management interface that met all our requirements. The solution was delivered within impressive timescales and at a competitive price. He kept us up to date throughout the project, was happy to be involved in conference calls to discuss setup and deployment, and was always incredibly responsive to our questions and requests for further enhancements. It was a pleasure working with Alastair and we look forward to the opportunity of working with him again in the future.

Stephen Connolly, Technical Team Manager
Parker Hannifin - Hemel Hempstead, Hertfordshire