Pavel Shipilov

CAPTCHAs lösen via Maschine Learning

/in , , , , , /by

Wie weit ist das maschinelle Lernen auf dem Gebiet der CAPTCHA-Lösung fortgeschritten?

Maschinelles Lernen ist mehr als ein Buzzword, denn unter der Haube stecken viele Algorithemen, die eine ganze Reihe von Problemen lösen können. Die Lösung von CAPTCHA ist dabei nur eine von vielen Aufgaben, die Machine Learning bewältigen kann. Durch die Arbeit an ein paar Problemen im Zusammenhang mit dem konvolutionellen neuronalen Netz haben wir festgestellt, dass es in diesem Bereich noch viel Verbesserungspotenzial gibt. Die Genauigkeit der Erkennung ist oftmals noch nicht gut genug. Schauen wir uns im Einzelnen an, welche Dienste wir haben, um dieses Problem anzugehen, und welche sich dabei als die besten erweisen.

Was ist CAPTCHA?

CAPTCHA ist kein fremder Begriff mehr für Web-Benutzer. Es handelt sich um die ärgerliche menschliche Validierungsprüfung, die auf vielen Websites hinzugefügt wird. Es ist ein Akronym für Completely Automated Public Turing test for tell Computer and Humans Apart. CAPTCHA kann als ein Computerprogramm bezeichnet werden, das dazu entwickelt wurde, Mensch und Maschine zu unterscheiden, um jede Art von illegaler Aktivität auf Websites zu verhindern. Der Sinn von CAPTCHA ist, dass nur ein Mensch diesen Test bestehen können sollte und Bots bzw. irgend eine Form automatisierter Skripte daran versagen. So entsteht ein Wettlauf zwischen CAPTCHA-Anbietern und Hacker-Lösungen, die auf den Einsatz von selbstlernenden Systemen setzen.

Warum müssen wir CAPTCHA lösen?

Heutzutage verwenden die Benutzer automatisierte CAPTCHA-Lösungen für verschiedene Anwendungsfälle. Und hier ein entscheidender Hinweis: Ähnlich wie Penetrationstesting ist der Einsatz gegen Dritte ohne vorherige Genehmigung illegal. Gegen eigene Anwendungen oder gegen Genehmigung (z. B. im Rahmen eines IT-Security-Tests) ist die Anwendung erlaubt. Hacker und Spammer verwenden die CAPTCHA-Bewältigung, um die E-Mail-Adressen der Benutzer zu erhalten, damit sie so viele Spams wie möglich erzeugen können oder um Bruteforce-Attacken durchführen zu können. Die legitimen Beispiele sind Fälle, in denen ein neuer Kunde oder Geschäftspartner zu Ihnen gekommen ist und Zugang zu Ihrer Programmierschnittstelle (API) benötigt, die noch nicht fertig ist oder nicht mit Ihnen geteilt werden kann, wegen eines Sicherheitsproblems oder Missbrauchs, den es verursachen könnte.

Für diese Anwendungsfälle sollen automatisierte Skripte CAPTCHA lösen. Es gibt verschiedene Arten von CAPTCHA: Textbasierte und bildbasierte CAPTCHA, reCAPTCHA und mathematisches CAPTCHA.

Es gibt einen Wettlauf zwischen CAPTCHA-Anbieter und automatisierten Lösungsversuchen. Die in CAPTCHA und reCAPTCHA verwendete Technologie werden deswegen immer intelligenter wird und Aktualisierungen der Zugangsmethoden häufiger. Das Aufrüsten hat begonnen.

Populäre Methoden für die CAPTCHA-Lösung

Die folgenden CAPTCHA-Lösungsmethoden stehen den Benutzern zur Lösung von CAPTCHA und reCAPTCHA zur Verfügung:

  1. OCR (optische Zeichenerkennung) via aktivierte Bots – Dieser spezielle Ansatz löst CAPTCHAs automatisch mit Hilfe der OCR-Technik (Optical Character Recognition). Werkzeuge wie Ocrad, tesseract lösen CAPTCHAs, aber mit sehr geringer Genauigkeit.
  2. Maschinenlernen — Unter Verwendung von Computer Vision, konvolutionalem neuronalem Netzwerk und Python-Frameworks und Bibliotheken wie Keras mit Tensorflow. Wir können tiefe neuronale Konvolutionsnetzmodelle trainieren, um die Buchstaben und Ziffern im CAPTCHA-Bild zu finden.
  3. Online-CAPTCHA-Lösungsdienstleistungen — Diese Dienste verfügen teilweise über menschliche Mitarbeiter, die ständig online verfügbar sind, um CAPTCHAs zu lösen. Wenn Sie Ihre CAPTCHA-Lösungsanfrage senden, übermittelt der Dienst sie an die Lösungsanbieter, die sie lösen und die Lösungen zurückschicken.

Leistungsanalyse der OCR-basierten Lösung

OCR Die OCR ist zwar eine kostengünstige Lösung, wenn es darum geht, eine große Anzahl von trivialen CAPTCHAs zu lösen, aber dennoch liefert sie nicht die erforderliche Genauigkeit. OCR-basierte Lösungen sind nach der Veröffentlichung von ReCaptcha V3 durch Google selten geworden. OCR-fähige Bots sind daher nicht dazu geeignet, CAPTCHA zu umgehen, die von Titanen wie Google, Facebook oder Twitter eingesetzt werden. Hierfür müsste ein besser ausgestattetes CAPTCHA-Lösungssystem eingesetzt werden.

OCR-basierte Lösungen lösen 1 aus 3 trivialen CAPTCHAs korrekt.

Leistungsanalyse der ML-basierten Methode

Schauen wir uns an, wie Lösungen auf dem Prinzip des Maschinenlernens funktionieren:

Die ML-basierte Verfahren verwenden OpenCV, um Konturen in einem Bild zu finden, das die durchgehenden Gebiete feststellt. Die Bilder werden mit der Technik der Schwellenwertbildung vorverarbeitet. Alle Bilder werden in Schwarzweiß konvertiert. Wir teilen das CAPTCHA-Bild mit der OpenCV-Funktion findContour() in verschiedene Buchstaben auf. Die verarbeiteten Bilder sind jetzt nur noch einzelne Buchstaben und Ziffern. Diese werden dann dem CNN-Modell zugeführt, um es zu trainieren. Und das trainierte CNN-Modell ist bereit, die richtige Captchas zu lösen.

Die Präzision einer solchen Lösung ist für alle textbasierten CAPTCHAs weitaus besser als die OCR-Lösung. Es gibt auch viele Nachteile dieser Lösung, denn sie löst nur eine bestimmte Art von CAPTCHAs und Google aktualisiert ständig seinen reCAPTCHA-Generierungsalgorithmus. Die letzte Aktualisierung schien die beste ReCaptcha-Aktualisierung zu sein, die disen Dienst bisher beeinflusst hat: Die regelmäßigen Nutzer hatten dabei kaum eine Veränderung der Schwierigkeit gespürt, während automatisierte Lösungen entweder gar nicht oder nur sehr langsam bzw. inakkurat funktionierten.

Das Modell wurde mit 1⁰⁴ Iterationen mit korrekten und zufälligen Stichproben und 1⁰⁵ Testbildern trainiert, und so wurde eine mittlere Genauigkeit von ~60% erreicht.

Bild-Quelle: “CAPTCHA Recognition with Active Deep Learning” @ TU München https://www.researchgate.net/publication/301620459_CAPTCHA_Recognition_with_Active_Deep_Learning

Wenn Ihr Anwendungsfall also darin besteht, eine Art von CAPTCHA mit ziemlich einfacher Komplexität zu lösen, können Sie ein solches trainiertes ML-Modell hervorragend nutzen. Eine bessere Captcha-Lösungslösung als OCR, muss aber noch eine ganze Menge Bereiche umfassen, um die Genauigkeit der Lösung zu gewährleisten.

Online-Captcha-Lösungsdienst

Online-CAPTCHA-Lösungsdienste sind bisher die bestmögliche Lösung für dieses Problem. Sie verfolgen alle Aktualisierungen von reCAPTCHA durch Google und bieten eine tadellose Genauigkeit von 99%.

Warum sind Online-Anti-Captcha-Dienste leistungsfähiger als andere Methoden?

Die OCR-basierten und ML-Lösungen weisen nach den bisherigen Forschungsarbeiten und Weiterentwicklungen viele Nachteile auf. Sie können nur triviale CAPTCHAs ohne wesentliche Genauigkeit lösen. Hier sind einige Punkte, die in diesem Zusammenhang zu berücksichtigen sind:

– Ein höherer Prozentsatz an korrekten Lösungen (OCR gibt bei wirklich komplizierten CAPTCHAs ein extrem hohes Maß an falschen Antworten; ganz zu schweigen davon, dass einige Arten von CAPTCHA überhaupt nicht mit OCR gelöst werden können, zumindest vorerst).

– Kontinuierlich fehlerfreie Arbeit ohne Unterbrechungen mit schneller Anpassung an die neu hinzugekommene Komplexität.

– Kostengünstig mit begrenzten Ressourcen und geringen Wartungskosten, da es keine Software- oder Hardwareprobleme gibt; alles, was Sie benötigen, ist eine Internetverbindung, um einfache Aufträge über die API des Anti-Captcha-Dienstes zu senden.

Die großen Anbieter von Online-Lösungsdiensten

Jetzt, nachdem wir die bessere Technik zur Lösung Ihrer CAPTCHAs geklärt haben, wollen wir unter allen Anti-Captcha-Diensten den besten auswählen. Einige Dienste bieten eine hohe Genauigkeit der Lösungen, API-Unterstützung für die Automatisierung und schnelle Antworten auf unsere Anfragen. Dazu gehören Dienste wie 2captcha, Imagetyperz, CaptchaSniper, etc.

2CAPTCHA ist einer der Dienste, die auf die Kombination von Machine Learning und echten Menschen setzen, um CAPTCHA zuverlässig zu lösen. Dabei versprechen Dienste wie 2captcha:

  • Schnelle Lösung mit 17 Sekunden für grafische und textuelle Captchas und ~23 Sekunden für ReCaptcha
  • Unterstützt alle populären Programmiersprachen mit einer umfassenden Dokumentation der fertigen Bibliotheken.
  • Hohe Genauigkeit (bis zu 99% je nach dem CAPTCHA-Typ).
  • Das Geld wird bei falschen Antworten zurückerstattet.
  • Fähigkeit, eine große Anzahl von Captchas zu lösen (mehr als 10.000 pro Minute)

Schlussfolgerung

Convolutional Neural Networks (CNN) wissen, wie die einfachsten Arten von Captcha zu bewältigen sind und werden auch mit der weiteren Enwicklung schritthalten können. Wir haben es mit einem Wettlauf um verkomplizierte CAPTCHAs und immer fähigeren Lösungen der automatisierten Erkennung zutun. Zur Zeit werden Online-Anti-Captcha-Dienste, die auf einen Mix aus maschinellem Lernen und menschlicher Intelligenz setzen, diesen Lösungen vorerst voraus sein.

Michael Lyamm

Top influential AI skills to target in 2020

/in , /by

Artificial intelligence in 2020: a trustable year!

The AI market is deemed to reach USD70 billion in the present year, thus causing a drastic effect on the government market, consumer, enterprise globally. 

With market uproars, it is certain obstacles are bound to intercept. However, the effects of artificial intelligence are poised to have huge potential in democratizing expensive services, or boost poor customer service, or assist during medical breakthroughs, and even lightened the work of the overburdened workforce. 

If you’re a tech optimist who believes in creating a world where man and machine can come closer and work together with humans, then you better need to possess mandatory AI skills. 

Based on this report, “2020 Workplace Learning Trends Report: The Skills of the Future,” you will come across how AI is reshaping the world and what are the skills that are a must-know for upcoming AI professionals. 

The report also states that the investment funds that are managed by AI are accounted for 35% of the stock market in America.

Finance machines will rise says a recent article in ‘The Economist.’ 

Trending by the industry’s insights and job trends, these are the skills AI professionals need to master in 2020:

  • Machine learning 

People often use the terms artificial intelligence and machine learning interchangeably, both being entirely different from each other. There’s a lot of confusion between both these terms, however, let us briefly understand what exactly is machine learning. 

Machine learning uses algorithms that obtain knowledge and skill through experience and without human intervention. It relies on big data sets that remind the data to find common patterns. 

For instance, if you provide machine learning programs with a lot of data on skin conditions and tell them what these conditions signify, these algorithms can easily mine the images (data) and help in analyzing the skin conditions even in the future. 

Now the algorithms will compare the images with the previous image data and try to identify the pattern that exists between them that have similar kind of patterns. 

However, if the algorithm is given a new image of a skin condition and the condition is unknown in the future as well. The algorithm will take the image to analyze it with the past and the present situation, thus the prediction of the condition will remain improper since one needs to feed new data for the algorithm to learn in order to predict what the condition is. However, AI will only learn by obtaining knowledge and learn how to apply these knowledges. Artificial intelligence helps to train computers aiming at a result that could provide a better outcome than a human can do. 

 

  • Python 

As the phrase is used, ‘orange is the new black’ so are AI and ML becoming the new black of the IT industry. With the extensive expansion of volumes of data, machine learning and artificial intelligence are used progressively for processing as well as analysis. To be honest, a human brain can function and analyze huge amounts of data but a human brain is restricted by limitations and can only contain a certain amount of processing or analyzing. 

AI in that aspect has a huge capacity and requires no limitation. 

Now Python plays a major impact on AI and machine learning as we undergo an upsurge for AI engineers. 

According to the latest trend search on Indeed, Python is said to be the most popular language for AI and ML, as stated by Jean Francois Puget, from the machine learning department in IBM. 

 

The how’s and why’s?

  • Python offers a low entry barrier for data scientists to effectively use Python without wasting much effort in learning the language. 
  • It has a great choice of library ecosystem, no wonder why everybody loves Python. For instance, Pandas for producing high-level data structure or data analysis, Matplotlib to create 2D plots and histograms, or Keras that is widely used for deep learning. 
  • Python as a language is flexible and is a great choice for machine learning. AI professionals can easily use Python along with other programming languages to achieve their target. 
  • Python offers easy readability for every developer, thus making it easy to understand the codes of their senior, change or develop a new one whenever needed. 

React (web)

If you’re a web developer thriving to enter the AI world, here’s a great chance for you. You can now build sweet AI apps using React.js. These offers web developers a new platform to bridge the gap between web developers and professionals who are getting trained in AI skills. 

A web developer can now easily build apps leveraging artificial intelligence that can learn from experiences or learns to react to the user’s inputs like facial expression, etc. 

Angular

If one isn’t aware, Google AI is build using Angular. Building a chatbot from scratch using a Dialog flow conversation platform, previously known as API.ai can be challenging. NLP (natural language processing) could be tough to deal with in machine learning. 

Docker 

Docker is now used in every field of the software industry. Confused with the term? Well, Docker is nothing but a DevOps tool. 

Docker can also be called as an exploiter tool exploiting operating-system-level virtualization that basically helps in developing and delivering software in packages that are called containers. 

Though this may sound complicated, you need to simply know that Docker is a complete environment that offers you a platform to build as well as deploy software. 

In a nutshell, one may say that Docker can be employed for varied phases of the machine learning development cycle like data aggregation, data-gathering, model training, predictive analysis, and application deployment, etc. 

Tech professionals such as DevOps engineers and AI engineers will need to supercharge their skill set in 2020. Till the time this knowledge gap persists, we will continue to see talent shortage in the job market.

 

netapp

Wie künstliche Intelligenz hilft, Krankheiten zu bekämpfen

/in , , /by

Die Herausforderungen im Gesundheitswesen sind gewaltig. Die Kosten steigen, das Geld ist knapp und die Margen sinken. Gleichzeitig fehlt es an Pflegepersonal, die vorhandenen Fachkräfte sind überarbeitet. Kliniken müssen effizient wirtschaften, um Patienten die bestmögliche Versorgung zu gewährleisten – und gleichzeitig Datensicherheits- und Compliance-Anforderungen bei der Verarbeitung der anfallenden Daten sicherstellen.

Der Einsatz von künstlicher Intelligenz (KI) kann dabei helfen, dieses Dilemma zu lösen. Algorithmen werden immer besser in dem, was sie tun – und sie arbeiten exakt, schnell und günstig. KI unterstützt in der Medizin und Forschung dabei, Patienten besser zu versorgen, indem beispielsweise Krankheiten früher erkannt werden können. Mit ihrer Hilfe könnten unter anderem die Gesundheitsausgaben in Europa in den kommenden zehn Jahren um einen dreistelligen Milliardenbetrag gesenkt werden, so das Ergebnis der PwC-Studie „Sherlock in Health – How artificial intelligence may improve quality and efficiency, whilst reducing healthcare costs in Europe“. Des Weiteren haben die meisten Patienten keine Berührungsängste: 54 Prozent wären demnach schon heute bereit, sich auf KI und Robotik in der Medizin einzulassen.

KI, ML und DL als medizinische Unterstützung

Algorithmen können in der Medizin auf unterschiedliche Weisen genutzt werden. KI beschäftigt sich mit Methoden, bei denen Computertechnologien es ermöglichen, menschliches Verhalten zu imitieren. Im Rahmen der medizinischen Bildgebung kann KI beispielsweise schnell Anomalien identifizieren, die für das menschliche Auge zu winzig erscheinen – oder große Datenmengen durchforsten. Ein Computertomograph erzeugt bis zu 640 Schnittbilder bei einem einzigen Scan. Wenn ein Radiologe sie ansehen und bewerten müsste, wäre das mit einem sehr hohen Zeitaufwand verbunden. Eine spezielle KI-Applikation kann die Bilder dagegen schnell analysieren und diejenigen markieren, welche Anomalien aufweisen. Die Radiologen können sich damit auf ihre Hauptaufgaben konzentrieren – Diagnose und Heilung. 

Ein weiteres Anwendungsgebiet von künstlicher Intelligenz innerhalb der Medizin ist der Einsatz von Intelligent Agents (IA), zum Beispiel für die Überwachung von Vitalwerten von Patienten oder als Kontrollmechanismus für die Arbeit des Pflegepersonals, der Ärzte oder Apotheker. Die Anwendungen überprüfen dann automatisch, ob die verschriebenen Medikamente und Therapien zum Krankheitsbild und zu den Werten des Patienten passen. 

Anwendungen aus dem Teilbereich der KI „Maschinelles Lernen (ML)“ lernen eigenständig dazu, je mehr Daten erfasst werden. Chirurgen können ML beispielsweise als Unterstützung verwenden, um den richtigen orthopädischen Eingriff nach einer Sportverletzung vorzubereiten. Die Technologie analysiert Patientendaten und kann die Unterschiede bei Knieverletzungen unterschiedlicher Sportarten sichtbar machen. So stehen dem Arzt detaillierte Informationen zur Verfügung, auf deren Basis er die Behandlung fortsetzt.

Deep Learning (DL) wiederum gilt als Teilbereich von ML und geht noch einen Schritt weiter, indem die Algorithmen nicht nur in der Lage sind, selbständig dazuzulernen, sondern sich auch kontinuierlich zu verbessern. So werden große Datenmengen verarbeitet, aus denen Wissenschaftler mögliche neue Erkenntnisse für Behandlungserfolge gewinnen können. Mit Hilfe von DL lassen sich beispielsweise bisher unbekannte Verbindungen zwischen bestimmten demografischen oder genetischen Indikatoren und unterschiedlichen Krankheiten aufdecken. Ebenso gibt es DL-Algorithmen, die mithilfe großer Datenmengen so trainiert werden, dass sie kleinste Veränderungen in der Zellstruktur erkennen können, die beispielsweise zu Brustkrebs führen. Die Fehlerquote bei der Brustkrebserkennung kann damit um bis zu 85 Prozent sinken, so eine Untersuchung von NVIDIA.

Komplexe KI-Projekte benötigen eine passende IT-Infrastruktur

Damit KI, ML und DL im Gesundheitswesen effektiv eingesetzt werden können, gibt es einige Grundvoraussetzungen. Zunächst einmal sind große Datenbestände notwendig. Diese werden genutzt, um die Algorithmen zu trainieren, damit sie akkurat und autonom arbeiten sowie Muster erkennen und genaue Vorhersagen treffen können. Dabei gilt es so viele Daten wie möglich zu berücksichtigen, unabhängig ob sie über verschiedene Systeme verteilt sind, aus unterschiedlichen Quellen stammen oder von mehreren unterschiedlichen Sensoren gesammelt wurden. Jedoch sollten sie eine hohe Datenqualität aufweisen. Darüber hinaus kann es sich um verschiedene Typen von Daten handeln (strukturiert, semi-strukturiert, unstrukturiert), die sich dynamisch entwickeln und verändern. 

Damit Daten überall dort verfügbar sind, wo sie gebraucht werden, gilt es Datensilos aufzulösen und miteinander zu verknüpfen. KI-Projekte stellen somit hohe Anforderungen an die IT-Infrastruktur hinsichtlich Datenverfügbarkeit und Datenqualität, Skalierbarkeit, Informationssicherheit sowie Genauigkeit mit hohen Ansprüchen an die Performance. Eine solch komplexe KI-Umgebung selbst zu planen, zu implementieren und zu unterhalten stellt viele Unternehmen vor große Herausforderungen. Es gibt jedoch schon umfassende Lösungen am Markt. Beispielsweise bietet NetApp zusammen mit NVIDIA und Cisco eine Lösung für die genannten Herausforderungen: ONTAP AI. Diese Architektur vereinfacht das komplette Datenmanagement: Informationen werden über das System sicher erfasst, weitergeleitet und verarbeitet, zum Training verwendet und analysiert.

ONTAP AI basiert auf einer verifizierten Architektur, die NVIDIA DGX-1 GPU‘s mit NetApp All Flash FAS Storage und Cisco Netzwerken zusammenführt und die Effizienz Ihrer KI-/DL-Umgebungen steigert. Herzstück von ONTAP AI ist die NVIDIA DGX-1, ein vollständig integriertes und sofort einsatzbereites Hardware- und Softwaresystem, das speziell für DL entwickelt wurde. Die DGX Plattform nutzt den Deep-Learning-Software-Stack der NVIDIA GPU Cloud, der für maximale GPU-beschleunigte DL-Performance optimiert wurde. Mit dem schnellen All-Flash-Storage und den weltweit ersten End-to-End NVMe-Technologien sorgen NetApp All Flash FAS Systeme für einen kontinuierlichen Datenfluss. So wird sichergestellt, dass die DGX-GPUs optimal mit Daten zur Verarbeitung versorgt werden und somit ein Flaschenhals hinsichtlich Datenbereitstellung durch die Storage-Systeme vermieden wird. 

Schnelle Diagnose

ONTAP AI kommt beispielsweise bei „BacillAi“ zum Einsatz, einem System zur Behandlung von Tuberkulose der Technologieberatungsfirma Cambridge Consultants. Tuberkulose ist die zweithäufigste Todesursache in Entwicklungsländern, da die Krankheit mit einer aufwendigen Diagnose verbunden ist: Zellen einer Speichelprobe müssen unter dem Mikroskop gezählt werden. Doch dafür gibt es nur wenig ausgebildetes medizinisches Personal. BacillAi vereinfacht diesen Schritt – und liefert zudem genauere und schnellere Ergebnisse. Ein Smartphone nimmt die Bilder der Speichelprobe von einem Standardmikroskop auf. Der DL-Algorithmus identifiziert Tuberkulose-Zellen, zählt sie und bestimmt das Stadium der Krankheit. Die Diagnose erhält der medizinische Mitarbeiter über eine App – somit ist das Ergebnis der Untersuchung zudem digitalisiert.

Fazit 

Künstliche Intelligenz kann das Gesundheitswesen revolutionieren. Unternehmen müssen dafür große Datenmengen aus unterschiedlichen Quellen erfassen, vorbereiten, verschieben, auf sie zugreifen und sie schützen. Damit KI, ML und DL-Projekte erfolgreich sind, brauchen Unternehmen aber eine effiziente Daten-Pipeline und eine Architektur, die eine hohe Performance, Sicherheit und Skalierbarkeit ermöglicht.

Weiter Informationen zum Thema finden Sie HIER.

Simon Krannig

Daten als Frühwarnsystem einsetzen

/in , , , , /by

In der klassischen Business Intelligence haben Unternehmen jahrelang Daten in Data Warehouses gesammelt und analysiert, um aus der Vergangenheit Lehren für die Zukunft zu ziehen. Zu seiner Zeit war das eine Revolution, aber da es sich dabei vor allem um Daten aus Transaktionssystemen handelte, war der Nutzen begrenzt. Erst mit der Verbreitung des IoT und von Sensoren, die permanent Daten liefern, konnten auch Gründe für Fehler oder Maschinenausfälle ausgelesen werden. Und wenn diese Gründe bestimmten Mustern folgen, liegt es nahe, einzugreifen, bevor ein Problem auftritt – das ist der Grundgedanke hinter dem Konzept von Predictive Analytics.

Großes bisher meist ungenutztes Potential

Systeme, die Risiken und Abweichungen als Frühwarnsystem erkennen, besitzen ein enormes wirtschaftliches Potential. In der Produktion beispielsweise können Maschinen länger reibungsfrei laufen und auch die IT-Infrastruktur profitiert. Predictive Analytics verändern aber auch die Unternehmensführung von Grund auf: Wenn Entscheidungen nur noch auf Basis von Daten anstatt von einem „Bauchgefühl” getroffen werden, verändert sich auch das Machtgefälle zugunsten der IT.

Wenn Entscheider sich nur noch auf Daten verlassen sollen/wollen und ihr Bauchgefühl ausschalten müssen, dann führt das zu einer Art “kultureller Überforderung” wie die Studie „Predictive Analytics 2018“ von IDG Research Services zeigt. Aber den meisten von ihnen ist klar, dass an dem Thema auf lange Sicht kein Weg vorbei führt. Zum Zeitpunkt der Befragung, die schon etwas zurückliegt, stuften bereits 47 Prozent der Unternehmen die Relevanz von Predictive Analytics als sehr hoch (18 Prozent) oder hoch (29 Prozent) ein. Über ein Drittel war aber bereits der Überzeugung, das Predictive Analytics spätestens 2021 eine sehr wichtige Rolle spielen wird.

Intelligenz in den Workflow bringen

Für Managed-Cloud-Unternehmen wie Adacor gewinnt Predictive Analytics in zweierlei Hinsicht an Bedeutung. Zum einen lassen sich damit Prozesse verbessern, mit denen bereits in der Vergangenheit Themen wie das Management von Server-Log-Daten oder CPU-Auslastungen automatisiert und vorausschauend gesteuert wurden.

Für Private Cloud Services, der maßgeschneiderten Erweiterung von internen Rechenzentren bedeutet dies, Teile des Live-Monitorings nach und nach in ein Predictive-Monitoring umzuwandeln und so auf mögliche Ausfälle oder Beeinträchtigungen von Servern im Vorfeld zu reagieren, um so auch den Ausfall für den Kunden zu verhindern. In einem einfachen Beispiel bewertet ein Deep-Learning Modell, ob auf einem beliebigen System die Festplattenfüllstände in der Zukunft stabil verlaufen werden oder ob mit instabilem Verhalten zu rechnen ist. Wird Stabilität erwartet, dann kann ein simpleres Vorhersagemodell diese Stabilität nutzen und die Füllstände vorhersagen. Ist mit instabilem Verhalten zu rechnen, dann wissen die Administratoren, dass sie ein besonderes Auge auf das entsprechende System werfen sollten. So wird durch vergleichsweise einfache Predictive-Monitoring Methoden bereits eine deutlich erhöhte Ausfallsicherheit der Systeme gewährleistet.

Neben stark individualisierten Cloud-Lösungen werden in Zukunft standardisierte Angebote immer mehr nachgefragt werden, die durch Predictive-Analytic-Tools „intelligenter” werden. Übersetzt bedeutet das, maschinelles Lernen nach Möglichkeit automatisch auf neue Prozesse anzuwenden und so Server bzw. die Cloud noch leistungsfähiger und sicherer zu machen.

Size matters

Die Studie zeigte, dass vor allem große Unternehmen Ressourcen für Analytics-Projekte bereitstellen. Über ein Drittel von ihnen hatte bereits Analytics-Projekte umgesetzt, mehr als die Hälfte davon im Bereich Predictive Analytics. Kleine und mittelständische Firmen hingegen verfügten noch wenig über umfangreiche Analytics-Systeme. Die Ergebnisse aus den Predictive-Analytics-Projekten beeinflussen im Wesentlichen auch die Management-Entscheidungen. 94 Prozent der Firmen, die Predictive Analytics anwenden, steuern über die Auswertungen Prozesse vor allem im IT-Bereich, im strategischen Management sowie in Produktion und Fertigung. Die großen Unternehmen sind also größtenteils schon dabei, sich die Vorteile zu nutzen zu machen. Bei mittelständischen und kleineren Unternehmen besteht noch deutlicher Nachholbedarf. Schon die technischen Voraussetzungen genügen häufig nicht den Anforderungen.

Fast alle Branchen können profitieren

Das erstaunt, denn Predictive Analytics kann in vielen Bereichen als eine Art Frühwarnsystem eingesetzt werden. Es hilft nicht nur dabei, Maschinenausfälle bei Produktionsunternehmen durch vorausschauende Wartungen zu minimieren. Es kann zum Beispiel auch den Vertrieb von Handelsunternehmen optimieren. In der Medizin kommen bereits Methoden zum Einsatz, durch die sich Risikofaktoren schneller identifizieren und die Behandlung von Krankheiten insgesamt verbessern lässt. Versicherungen und auch Finanzinstitute kalkulieren ihre Produkte und Prämien seit jeher erfolgreich auf Basis von Wahrscheinlichkeitsanalysen und Hochrechnungen. Auch im Bereich der Betrugsprävention werden entsprechend Methoden und Tools verstärkt eingesetzt, um Kriminellen das Handwerk zu legen.

Man sieht, es lohnt sich für Unternehmen, die Daten sammeln, ihre Strategie an die neuen Technologien anzupassen. Die aktuellen Möglichkeiten zur Analyse und Aggregierung von Daten und Informationen sind extrem groß. Es kommt darauf an, Muster in den „Big Data” zu erkennen und diese richtig zu interpretieren – anstatt dieselben Fehler immer und immer wieder zu machen.

“Saubere Ablage“ bringt Unternehmen nicht weiter

Unternehmen, die Daten sammeln, um diese lediglich sauber abzulegen und zu archivieren, sollten Ihre Strategie an die neuen Möglichkeiten des Predictive Analytics anpassen. Die aktuellen Möglichkeiten zur Auswertung und zur Verdichtung von Daten zu Informationen und somit zur Generierung von Wissen sind extrem groß. Nur wer Muster im großen Reich der Daten erkennt und diese auch richtig interpretieren kann, wird kann mit Predicitve Analytics ein Frühwarnsystem zu seinen Gunsten aufbauen.

Manoj Rupareliya

AI For Advertisers: How Data Analytics Can Change The Maths Of Advertising?

/in , , , , /by

All Images Credit: Freepik

The task of understanding a customer’s journey and designing your marketing strategy accordingly can be difficult in this data-driven world. Today, the customer expresses their needs in myriad forms of requests.

Consumers express their needs and want attitudes, and values in various forms through search, comments, blogs, Tweets, “likes,” videos, and conversations and access such data across many channels like web, mobile, and face to face. Volume, variety, velocity and veracity of the data accumulated through these customer interactions are huge.

BigData and data analytics can be leveraged to understand several phases of the customer journey. There are risks involved in using Artificial Intelligence for the marketing data analysis of data breach and even manipulation. But, AI do have brighter prospects when it comes to marketing and advertiser applications.

As the CEO of a technology firm Chop Dawg and marketer, Joshua Davidson puts it, “AI-powered apps are going to be the future for us, and there are several industries that are ripe for this.” The mobile-first strategy of many enterprises has powered the use of AI for digital marketing and developing technologies and innovations to power industries with intelligent systems.

How AI and Machine learning are affecting customer journeys?

Any consumer journey begins with the recognition of a problem and then stages like initial consideration, active evaluation, purchase, and postpurchase come through up till the consumer journey is over. The need for identifying the purchasing and need patterns of the consumers and finding the buyer personas to strategize the marketing for them.

Need and Want Recognition:

Identifying a need is quite difficult as it is the most initial level of a consumer’s journey and it is more on the category level than at a brand level. Marketers and advertisers are relying on techniques like market research, web analytics, and data mining to build consumer profiles and buyer’s persona for understanding the needs and influencing the purchase of products. AI can help identify these wants and needs in real-time as the consumers usually express their needs and wants online and help build profiles more quickly.

AI technologies offered by several firms help in consumer profiling. Firms like Microsoft offers Azure that crunches billions of data points in seconds to determine the needs of consumers. It then personalizes web content on specific platforms in real-time to align with those status-updates. Consumer digital footprints are evolving through social media status updates, purchasing behavior, online comments and posts. Ai tends to update these profiles continuously through machine learning techniques.

Initial Consideration:

A key objective of advertising is to insert a brand into the consideration set of the consumers when they are looking for deliberate offerings. Advertising includes increasing the visibility of brands and emphasize on the key reasons for consideration. Advertisers currently use search optimization, paid search advertisements, organic search, or advertisement retargeting for finding the consideration and increase the probability of consumer consideration.

AI can leverage machine learning and data analytics to help with search, identify and rank functions of consumer consideration that can match the real-time considerations at any specific time. Take an example of Google Adwords, it analyzes the consumer data and helps advertisers make clearer distinctions between qualified and unqualified leads for better targeting.

Google uses AI to analyze the search-query data by considering, not only the keywords but also context words and phrases, consumer activity data and other BigData. Then, Google identifies valuable subsets of consumers and more accurate targeting.

Active Evaluation: 

When consumers narrow it down to a few choices of brands, advertisers need to insert trust and value among the consumers for brands. A common technique is to identify the higher purchase consumers and persuade them through persuasive content and advertisement. AI can support these tasks using some techniques:

Predictive Lead Scoring: Predictive lead scoring by leveraging machine learning techniques of predictive analytics to allow marketers to make accurate predictions related to the intent of purchase for consumers. A machine learning algorithm runs through a database of existing consumer data, then recognize trends and patterns and after processing the external data on consumer activities and interests, creates robust consumer profiles for advertisers.

Natural Language Generation: By leveraging the image, speech recognition and natural language generation, machine learning enables marketers to curate content while learning from the consumer behavior in real-time scenarios and adjusts the content according to the profiles on the fly.

Emotion AI: Marketers use emotion AI to understand consumer sentiment and feel about the brand in general. By tapping into the reviews, blogs or videos they understand the mood of customers. Marketers also use emotion AI to pretest advertisements before its release. The famous example of Kelloggs, which used emotion AI to help devise an advertising campaign for their cereal, eliminating the advertisement executions whenever the consumer engagement dropped.

Purchase: 

As the consumers decide which brands to choose and what it’s worth, advertising aims to move them out of the decision process and push for the purchase by reinforcing the value of the brand compared with its competition.

Advertisers can insert such value by emphasizing convenience and information about where to buy the product, how to buy the product and reassuring the value through warranties and guarantees. Many marketers also emphasize on rapid return policies and purchase incentives.

AI can completely change the purchase process through dynamic pricing, which encompasses real-time price adjustments on the basis of information such as demand and other consumer-behavior variables, seasonality, and competitor activities.

Post-Purchase: 

Aftersales services can be improved through intelligent systems using AI technologies and machine learning techniques. Marketers and advertisers can hire dedicated developers to design intelligent virtual agents or chatbots that can reinforce the value and performance of a brand among consumers.

Marketers can leverage an intelligent technique known as Propensity modeling to identify the most valuable customers on the basis of lifetime value, likelihood of reengagement, propensity to churn, and other key performance measures of interest. Then advertisers can personalize their communication with these customers on the basis of these data.

Conclusion:

AI has shifted the focus of advertisers and marketers towards the customer-first strategies and enhanced the heuristics of customer engagement. Machine learning and IoT(Internet of Things) has already changed the way customer interact with the brands and this transition has come at a time when advertisers and marketers are looking for new ways to tap into the customer mindset and buyer’s persona.

All Images Credit: Freepik

Valentin Curtef

The importance of being Data Scientist

/in , , , , , , /by

Header-Image by Clint Adair on Unsplash.

The incredible results of Machine Learning and Artificial Intelligence, Deep Learning in particular, could give the impression that Data Scientist are like magician. Just think of it. Recognising faces of people, translating from one language to another, diagnosing diseases from images, computing which product should be shown for us next to buy and so on from numbers only. Numbers which existed for centuries. What a perfect illusion. But it is only an illusion, as Data Scientist existed as well for centuries. However, there is a difference between the one from today compared to the one from the past: evolution.

The main activity of Data Scientist is to work with information also called data. Records of data are as old as mankind, but only within the 16 century did it include also numeric forms — as numbers started to gain more and more ground developing their own symbols. Numerical data, from a given phenomenon — being an experiment or the counts of sheep sold by week over the year –, was from early on saved in tabular form. Such a way to record data is interlinked with the supposition that information can be extracted from it, that knowledge — in form of functions — is hidden and awaits to be discovered. Collecting data and determining the function best fitting them let scientist to new insight into the law of nature right away: Galileo’s velocity law, Kepler’s planetary law, Newton theory of gravity etc.

Such incredible results where not possible without the data. In the past, one was able to collect data only as a scientist, an academic. In many instances, one needed to perform the experiment by himself. Gathering data was tiresome and very time consuming. No sensor which automatically measures the temperature or humidity, no computer on which all the data are written with the corresponding time stamp and are immediately available to be analysed. No, everything was performed manually: from the collection of the data to the tiresome computation.

More then that. Just think of Michael Faraday and Hermann Hertz and there experiments. Such endeavour where what we will call today an one-man-show. Both of them developed parts of the needed physics and tools, detailed the needed experiment settings, conducting the experiment and collect the data and, finally, computing the results. The same is true for many other experiments of their time. In biology Charles Darwin makes its case regarding evolution from the data collected in his expeditions on board of the Beagle over a period of 5 years, or Gregor Mendel which carry out a study of pea regarding the inherence of traits. In physics Blaise Pascal used the barometer to determine the atmospheric pressure or in chemistry Antoine Lavoisier discovers from many reaction in closed container that the total mass does not change over time. In that age, one person was enough to perform everything and was the reason why the last part, of a data scientist, could not be thought of without the rest. It was inseparable from the rest of the phenomenon.

With the advance of technology, theory and experimental tools was a specialisation gradually inescapable. As the experiments grow more and more complex, the background and condition in which the experiments were performed grow more and more complex. Newton managed to make first observation on light with a simple prism, but observing the line and bands from the light of the sun more than a century and half later by Joseph von Fraunhofer was a different matter. The small improvements over the centuries culminated in experiments like CERN or the Human Genome Project which would be impossible to be carried out by one person alone. Not only was it necessary to assign a different person with special skills for a separate task or subtask, but entire teams. CERN employs today around 17 500 people. Only in such a line of specialisation can one concentrate only on one task alone. Thus, some will have just the knowledge about the theory, some just of the tools of the experiment, other just how to collect the data and, again, some other just how to analyse best the recorded data.

If there is a specialisation regarding every part of the experiment, what makes Data Scientist so special? It is impossible to validate a theory, deciding which market strategy is best without the work of the Data Scientist. It is the reason why one starts today recording data in the first place. Not only the size of the experiment has grown in the past centuries, but also the size of the data. Gauss manage to determine the orbit of Ceres with less than 20 measurements, whereas the new picture about the black hole took 5 petabytes of recorded data. To put this in perspective, 1.5 petabytes corresponds to 33 billion photos or 66.5 years of HD-TV videos. If one includes also the time to eat and sleep, than 5 petabytes would be enough for a life time.

For Faraday and Hertz, and all the other scientist of their time, the goal was to find some relationship in the scarce data they painstakingly recorded. Due to time limitations, no special skills could be developed regarding only the part of analysing data. Not only are Data Scientist better equipped as the scientist of the past in analysing data, but they managed to develop new methods like Deep Learning, which have no mathematical foundation yet in spate of their success. Data Scientist developed over the centuries to the seldom branch of science which bring together what the scientific specialisation was forced to split.

What was impossible to conceive in the 19 century, became more and more a reality at the end of the 20 century and developed to a stand alone discipline at the beginning of the 21 century. Such a development is not only natural, but also the ground for the development of A.I. in general. The mathematical tools needed for such an endeavour where already developed by the half of the 20 century in the period when computing power was scars. Although the mathematical methods were present for everyone, to understand them and learn how to apply them developed quite differently within every individual field in which Machine Learning/A.I. was applied. The way the same method would be applied by a physicist, a chemist, a biologist or an economist would differ so radical, that different words emerged which lead to different langues for similar algorithms. Even today, when Data Science has became a independent branch, two different Data Scientists from different application background could find it difficult to understand each other only from a language point of view. The moment they look at the methods and code the differences will slowly melt away.

Finding a universal language for Data Science is one of the next important steps in the development of A.I. Then it would be possible for a Data Scientist to successfully finish a project in industry, turn to a new one in physics, then biology and returning to industry without much need to learn special new languages in order to be able to perform each tasks. It would be possible to concentrate on that what a Data Scientist does best: find the best algorithm. In other words, a Data Scientist could resolve problems independent of the background the problem was stated.

This is the most important aspect that distinguish the Data Scientist. A mathematician is limited to solve problems in mathematics alone, a physicist is able to solve problems only in physics, a biologist problems only in biology. With a unique language regarding the methods and strategies to solve Machine Learning/A.I. problems, a Data Scientist can solve a problem independent of the field. Specialisation put different branches of science at drift from each other, but it is the evolution of the role of the Data Scientist to synthesize from all of them and find the quintessence in a language which transpire beyond all the field of science. The emerging language of Data Science is a new building block, a new mathematical language of nature.

Although such a perspective does not yet exists, the principal component of Machine Learning/A.I. already have such proprieties partially in form of data. Because predicting for example the numbers of eggs sold by a company or the numbers of patients which developed immune bacteria to a specific antibiotic in all hospital in a country can be performed by the same prediction method. The data do not carry any information about the entities which are being predicted. It does not matter anymore if the data are from Faraday’s experiment, CERN of Human Genome. The same data set and its corresponding prediction could stand literary for anything. Thus, the result of the prediction — what we would call for a human being intuition and/or estimation — would be independent of the domain, the area of knowledge it originated.

It also lies at the very heart of A.I., the dream of researcher to create self acting entities, that is machines with consciousness. This implies that the algorithms must be able to determine which task, model is relevant at a given moment. It would be to cumbersome to have a model for every task and and every field and then try to connect them all in one. The independence of scientific language, like of data, is thus a mandatory step. It also means that developing A.I. is not only connected to develop a new consciousness, but, and most important, to the development of our one.

Michael Lyamm

AI Experts: The Next Frontier in AI After the 2020 Job Crisis

/in , /by

Beware the perils of AI boom!

Isn’t this something that should ring alarm bells to upgrade your AI skills.

Artificial intelligence has grown smarter putting people in awe with a question, “Is my job safe?” Should we be afraid? It is but a simple question with a rather perplexing answer, I’m not skilled ready. Your view will depend on whether you’ll be able to develop skills that will surpass the redundant skills you possess today.No doubt, the AI domain is thriving and humans are scared. 

Even organizations such as McKinsey predicts the doom and gloom scenario where one-third of the workers’ jobs will be taken over due to automation by 2030.

In the next decade, AI and automation could banish 54 million Americans out of their workspace. With rapid technological growth, machines are now outperforming the number of tasks traditionally done by manpower.

What’s more?

  • Walmart has the fastest automated truck unloader that helps scan unloaded items on a priority basis. 
  • McDonald replaces drive-thru workers with order-taking AI and cashiers along with self-checkout kiosks. 
  • While farms in California hire robots to harvest lettuce. 

Fear facts appearing real

Near about 670,000 U.S. jobs were replaced between 1990 to 2007, mostly in the manufacturing sector. But this trend is already accelerating as it advances in mobile technology, data transfer, AI, and computing speed. 

On its face, jobs that involve physical tasks in predictable environments will be at higher risk. For instance, The Palm Beach County Court recently made use of four robots (Rosie Tobor, Kitt Robbie, Speedy, and Wally Bishop) to read out the court filings, input data into the case management system, and fill out docket sheets. Also, at certain places in China, waiters were being replaced with robots.

On the contrary, jobs that include creative thinking, social interaction, and managing people will barely involve automation.

Though you think your job is safe, it isn’t. 

History has warned us of the apocalyptic happenings about technology replacing our jobs. There has always been a difficult transition to jobs that require newer skillsets. McKinsey, in its study, mentioned 8-9% of 2030’s labor demand will be in newer job roles that did not exist before. 

AI to take over the world – or is it? 

There is still but a grim prognostic about the robot apocalypse. But it’s not the time to celebrate.

As warned by Russian president Vladimir Putin, “The nation that leads in AI will be the ruler of the world.”

Artificial intelligence is yet to replace the human workforce, but it is still considered an invaluable tool today.

According to Forrester, firms will now address the pragmatic side of AI about having a better understanding of the challenges faced, to embrace the idea which is, no pain means no AI gain. The AI reality is here, right now. Organizations have now realized what they can do and what they cannot. Their focus is now projected toward taking proactive measures to produce more AI talents like AI experts and AI specialists, etc. 

Is there a timeframe where AI will overtake the human race?

It is only a matter of time when artificial intelligence will become smarter than its human creators.

Experts have already started to build a world that is brimming with AI. But sadly, in the present, most individuals are yet to know what AI even is. By the next decade, AI is predicted to outperform human in multiple activities such as,

  • Translating languages – 2024
  • Writing high-school essays – 2025
  • Driving a truck – 2027
  • Working in the retail sector – 2031
  • Or writing a best-selling book – 2049
  • Work as a surgeon – 2053

Beyond the shadow of a doubt, as artificial intelligence continues to grow, some experts say we’ll eventually hit the plateau. On the research side, there will be a snowball of AI challenges. Therefore, to tackle these challenges, the demand for AI experts will dramatically upsurge.

In addition to the dearth of AI talent, the transition may bring new challenges both for policymakers as well as AI professionals. 

“High-level machine intelligence will start performing any task better than the humans by 2060, and will take away human jobs by 2136, predicted a study done by multiple researchers from Yale and Oxford University.”

To stay prepared for the upcoming challenges, upskilling is the right way to reshape and overcome the AI jobs crisis. 

Upskilling in AI is the new mandate

Notably, as AI takes on to become the next technology revolution, certifications in artificial intelligence will keep you one step ahead. 

The advent of artificial intelligence has advanced at a level where there is a dire need for AI engineers. Now is the right time to pursue a career in artificial intelligence.

The current job market is flooded with multiple AI career options, but there’s a significant dearth of talent in the AI field. Professionals like software engineers have an upper hand in the AI industry. Additional certification programs have the capability of boosting the credibility of such individuals. 

Just like any other technology predictions, it’s an ideal decision to take up AI certifications. Staying up-to-date will prevent you from unnecessary panic – where AI could help you and not hurt you.

An economist Yale Brozen from the University of Chicago found out about technology destroying approximately 12 million jobs in the 1950s. But consecutively it also created over 20 million jobs as vast productivity leading toward the demand for more workers to keep up the pace with the rising demand.

Do you still need a reason not to adopt AI?

The AI catastrophe that dooms us is a threat to humans today. The pronouncement has retreated into a grim future where ignorance is not the solution. 

The pervasive answer is, only individuals that can make progress in their AI career will make it through the job crisis. 

Do you think your job is safe? Think again!

Akshay Chavan

Visual Question Answering with Keras – Part 2: Making Computers Intelligent to answer from images

/in , , , , , , /by

Making Computers Intelligent to answer from images

This is my second blog on Visual Question Answering, in the last blog, I have introduced to VQA, available datasets and some of the real-life applications of VQA. If you have not gone through then I would highly recommend you to go through it. Click here for more details about it.

In this blog post, I will walk through the implementation of VQA in Keras.

You can download the dataset from here: https://visualqa.org/index.html. All my experiments were performed with VQA v2 and I have used a very tiny subset of entire dataset i.e all samples for training and testing from the validation set.

Table of contents:

  1. Preprocessing Data
  2. Process overview for VQA
  3. Data Preprocessing – Images
  4. Data Preprocessing through the spaCy library- Questions
  5. Model Architecture
  6. Defining model parameters
  7. Evaluating the model
  8. Final Thought
  9. References

NOTE: The purpose of this blog is not to get the state-of-art performance on VQA. But the idea is to get familiar with the concept. All my experiments were performed with the validation set only.

Full code on my Github here.


1. Preprocessing Data:

If you have downloaded the dataset then the question and answers (called as annotations) are in JSON format. I have provided the code to extract the questions, annotations and other useful information in my Github repository. All extracted information is stored in .txt file format. After executing code the preprocessing directory will have the following structure.

All text files will be used for training.

 

2. Process overview for VQA:

As we have discussed in previous post visual question answering is broken down into 2 broad-spectrum i.e. vision and text.  I will represent the Neural Network approach to this problem using the Convolutional Neural Network (for image data) and Recurrent Neural Network(for text data). 

If you are not familiar with RNN (more precisely LSTM) then I would highly recommend you to go through Colah’s blog and Andrej Karpathy blog. The concepts discussed in this blogs are extensively used in my post.

The main idea is to get features for images from CNN and features for the text from RNN and finally combine them to generate the answer by passing them through some fully connected layers. The below figure shows the same idea.

 

I have used VGG-16 to extract the features from the image and LSTM layers to extract the features from questions and combining them to get the answer.

3. Data Preprocessing – Images:

Images are nothing but one of the input to our model. But as you already may know that before feeding images to the model we need to convert into the fixed-size vector.

So we need to convert every image into a fixed-size vector then it can be fed to the neural network. For this, we will use the VGG-16 pretrained model. VGG-16 model architecture is trained on millions on the Imagenet dataset to classify the image into one of 1000 classes. Here our task is not to classify the image but to get the bottleneck features from the second last layer.

Hence after removing the softmax layer, we get a 4096-dimensional vector representation (bottleneck features) for each image.

Image Source: https://www.cs.toronto.edu/~frossard/post/vgg16/

 

For the VQA dataset, the images are from the COCO dataset and each image has unique id associated with it. All these images are passed through the VGG-16 architecture and their vector representation is stored in the “.mat” file along with id. So in actual, we need not have to implement VGG-16 architecture instead we just do look up into file with the id of the image at hand and we will get a 4096-dimensional vector representation for the image.

4. Data Preprocessing through the spaCy library- Questions:

spaCy is a free, open-source library for advanced Natural Language Processing (NLP) in Python. As we have converted images into a fixed 4096-dimensional vector we also need to convert questions into a fixed-size vector representation. For installing spaCy click here. 

You might know that for training word embeddings in Keras we have a layer called an Embedding layer which takes a word and embeds it into a higher dimensional vector representation. But by using the spaCy library we do not have to train the get the vector representation in higher dimensions.

 

This model is actually trained on billions of tokens of the large corpus. So we just need to call the vector method of spaCy class and will get vector representation for word.

After fitting, the vector method on tokens of each question will get the 300-dimensional fixed representation for each word.

5. Model Architecture:

In our problem the input consists of two parts i.e an image vector, and a question, we cannot use the Sequential API of the Keras library. For this reason, we use the Functional API which allows us to create multiple models and finally merge models.

The below picture shows the high-level architecture idea of submodules of neural network.

After concatenating the 2 different models the summary will look like the following.

The below plot helps us to visualize neural network architecture and to understand the two types of input:

 

6. Defining model parameters:

The hyperparameters that we are going to use for our model is defined as follows:

If you know what this parameter means then you can play around it and can get better results.

Time Taken: I used the GPU on https://colab.research.google.com and hence it took me approximately 2 hours to train the model for 5 epochs. However, if you train it on a PC without GPU, it could take more time depending on the configuration of your machine.

7. Evaluating the model:

Since I have used the very small dataset for performing these experiments I am not able to get very good accuracy. The below code will calculate the accuracy of the model.

 

Since I have trained a model multiple times with different parameters you will not get the same accuracy as me. If you want you can directly download mode.h5 file from my google drive.

 

8. Final Thoughts:

One of the interesting thing about VQA is that it a completely new field. So there is absolutely no end to what you can do to solve this problem. Below are some tips while replicating the code.

  1. Start with a very small subset of data: When you start implementing I suggest you start with a very small amount of data. Because once you are ready with the whole setup then you can scale it any time.
  2. Understand the code: Understanding code line by line is very much helpful to match your theoretical knowledge. So for that, I suggest you can take very few samples(maybe 20 or less) and run a small chunk (2 to 3 lines) of code to get the functionality of each part.
  3. Be patient: One of the mistakes that I did while starting with this project was to do everything at one go. If you get some error while replicating code spend 4 to 5 days harder on that. Even after that if you won’t able to solve, I would suggest you resume after a break of 1 or 2 days. 

VQA is the intersection of NLP and CV and hopefully, this project will give you a better understanding (more precisely practically) with most of the deep learning concepts.

If you want to improve the performance of the model below are few tips you can try:

  1. Use larger datasets
  2. Try Building more complex models like Attention, etc
  3. Try using other pre-trained word embeddings like Glove 
  4. Try using a different architecture 
  5. Do more hyperparameter tuning

The list is endless and it goes on.

In the blog, I have not provided the complete code you can get it from my Github repository.

9. References:

  1. https://blog.floydhub.com/asking-questions-to-images-with-deep-learning/
  2. https://tryolabs.com/blog/2018/03/01/introduction-to-visual-question-answering/
  3. https://github.com/sominwadhwa/vqamd_floyd

Haroon Bhutta

The Future of AI in Dental Technology

/in , , /by

As we develop more advanced technology, we begin to learn that artificial intelligence can have more and more of an impact on our lives and industries that we have gotten used to being the same over the past decades. One of those industries is dentistry. In your lifetime, you’ve probably not seen many changes in technology, but a boom around artificial intelligence and technology has opened the door for AI in dental technologies.

How Can AI Help?

Though dentists take a lot of pride in their craft and career, most acknowledge that AI can do some things that they can’t do or would make their job easier if they didn’t have to do. AI can perform a number of both simple and advanced tasks. Let’s take a look at some areas that many in the dental industry feel that AI can be of assistance.

Repetitive, Menial Tasks

The most obvious area that AI can help out when it comes to dentistry is with repetitive and menial simple tasks. There are many administrative tasks in the dentistry industry that can be sped up and made more cost-effective with the use of AI. If we can train a computer to do some of these tasks, we may be able to free up more time for our dentists to focus on more important matters and improve their job performance as well. One primary use of AI is virtual consultations that offices like Philly Braces are offering. This saves patients time when they come in as the Doctor already knows what the next steps in their treatment will be.

Using AI to do some basic computer tasks is already being done on a small scale by some, but we have yet to see a very large scale implementation of this technology. We would expect that to happen soon, with how promising and cost-effective the technology has proven to be.

Reducing Misdiagnosis

One area that many think that AI can help a lot in is misdiagnosis. Though dentists do their best, there is still a nearly 20% misdiagnosis rate when reading x-rays in dentistry. We like to think that a human can read an x-ray better, but this may not be the case. AI technology can certainly be trained to read an x-ray and there have been some trials to suggest that they can do it better and identify key conditions that we often misread.

A world with AI diagnosis that is accurate and quicker will save time, money, and lead to better dental health among patients. It hasn’t yet come to fruition, but this seems to be the next major step for AI in dentistry.

Artificial Intelligence Assistants

Once it has been demonstrated that AI can perform a range of tasks that are useful to dentists, the next logical step is to combine those skills to make a fully-functional AI dental assistant. A machine like this has not yet been developed, but we can imagine that it would be an interface that could be spoken to similar to Alexa. The dentist would request vital information and other health history data from a patient or set of patients to assist in the treatment process. This would undoubtedly be a huge step forward and bring a lot of computing power into the average dentist office.

Conclusion

It’s clear that AI has a bright future in the dental industry and has already shown some of the essential skills that it can help with in order to provide more comprehensive and accurate care to dental patients. Some offices like Westwood Orthodontics already use AI in the form of a virtual consult to diagnose issues and provide treatment options before patients actually step foot in the office. Though not nearly all applications that AI can provide have been explored, we are well on our way to discovering the vast benefits of artificial intelligence for both patients and practices in the dental healthcare industry.

Michael Lyamm

Accelerate your AI Skills Today: A Million Dollar Job!

/in , , , , , /by

The skyrocketing salaries ($1m per year) of AI engineers is not a hype. It is the fact of current corporate world, where you will witness a shift that is inevitable.

We’ve already set our feet at the edge of the technological revolution. A revolution that is at the verge of altering the way we live and work. As the fact suggests, humanity has fundamentally developed human production in three revolutions, and we’re now entering the fourth revolution. In its scope, the fourth revolution projects a transformation that is unlike anything we humans have ever experienced.

  • The first revolution had the world transformed from rural to urban
  • the emergence of mass production in the second revolution
  • third introduced the digital revolution
  • The fourth industrial revolution is anxious to integrate technologies into our lives.

And all thanks to artificial intelligence (AI). An advanced technology that surrounds us, from virtual assistants to software that translates to self-driving cars.

The rise of AI at an exponential rate has disrupted almost every industry. So much so that AI is being rated as one-million-dollar profession.

Did this grab your attention? It did?

Now, what if we were to tell you that the salary compensation for AI experts has grown dramatically. AI and machine learning are fields that have a mountain of demand in the tech industry today but has sparse supply.

AI field is growing at a quicker pace and salaries are skyrocketing! Read it for yourself to know what AI experts, AI researchers and any other AI talent are commanding today.

  • A top-class AI research laboratory, OpenAI says that techies in the AI field are projected to earn a salary compensation ranging between $300 to $500k for fresh graduates. However, expert professionals could earn anywhere up to $1m.
  • Whopping salary package of above 100 million yen that amounts to $1m is being offered to AI geniuses by a Japanese firm, Start Today. A firm that operates a fashion shopping website named Zozotown.

Does this leave you with a question – Is this a right opportunity for you to jump in the field and make hay while the sun is shining? 

And the answer to this question is – yes, it is the right opportunity for any developer seeking a role in the AI industry. It can be your chance to bridge the skill shortage in the AI field either by upskilling or reskilling yourself in the field of AI.

There are a wide varieties of roles available for an AI enthusiast like you. And certain areas are like AI Engineers and AI Researchers are high in demand, as there are not many professionals who have robust AI knowledge.

According to a job report, “The Future of Jobs 2018,” a prediction was made suggesting that machines and algorithms will create around 133 million new job roles by 2022.

AI and machine learning will dominate the tech world. The World Economic Forum says that several sectors have started embracing AI and machine learning to tackle challenges in certain fields such as advertising, supply chain, manufacturing, smart cities, drones, and cybersecurity.

Unraveling the AI realm

From chatbots to financial planners, AI is impacting the way businesses function on a day-today basis. AI makes the work simpler, as it provides variables, which makes the work more streamlined.

Alright! You know that

  • the demand for AI professionals is rising exponentially and that there is just a trickle of supply
  • the AI professionals are demanding skyrocketing salaries

However, beyond that how much more do you know about AI?

Considering the fact that our lives have already been touched by AI (think Alexa, and Siri), it is just a matter of time when AI will become an indispensable part of our lives.

As Gartner predicts that 2020 will be an important year for business growth in AI. Thus, it is possible to witness significant sparks for employment growth. Though AI predicts to diminish 1.8 million jobs, it is also said to replace it with 2.3 million jobs that will be created. As we look forward to stepping into 2020, AI-related job roles are set to make positive progress of achieving 2 million net-new employments by 2025.

With AI promising to score fat paychecks that would reach millions, AI experts are struggling to find new ways to pick up nouveau skills. However, one of the biggest impacts that affect the job market today is the scarcity of talent in this field.

The best way to stay relevant and employable in AI is probably by “reskilling,” and “upskilling.” And  AI certifications is considered ideal for those in the current workforce.

Looking to upskill yourself – here’s how you can become an AI engineer today.

Top three ways to enhance your artificial intelligence career:

  1. Acquire skills in Statistics and Machine Learning: If you’re getting into the field of machine learning, it is crucial that you have in-depth knowledge of statistics. Statistics is considered a prerequisite to the ML field. Both the fields are tightly related. Machine learning models are created to make accurate predictions while statistical models do the job of interpreting the relationship between variables. Many ML techniques heavily rely on the theory obtained through statistics. Thus, having extensive knowledge in statistics help initiate the first step towards an AI career.
  2. Online certification programs in AI skills: Opting for AI certifications will boost your credibility amongst potential employers. Certifications will also enhance your earning potential and increase your marketability. If you’re looking for a change and to be a part of something impactful; join the AI bandwagon. The IT industry is growing at breakneck speed; it is now that businesses are realizing how important it is to hire professionals with certain skillsets. Specifically, those who are certified in AI are becoming sought after in the job market.
  3. Hands-on experience: There’s a vast difference in theory and practical knowledge. One needs to familiarize themselves with the latest tools and technologies used by the industry. This is possible only if the individual is willing to work on projects and build things from scratch.

Despite all the promises, AI does prove to be a threat to job holders, if they don’t upskill or reskill themselves. The upcoming AI revolution will definitely disrupt the way we work, however, it will leave room for humans to perform more creative jobs in the future corporate world.

So a word of advice is to be prepared and stay future ready.