AI + Engineering: from Physical to Virtual Sensors using Deep Learning, Enrico Busto - Add-for
0 likes513 views
The document discusses the integration of deep learning algorithms into traditional engineering applications like signal filtering and control systems, enhancing functionality in industrial settings. It highlights various projects, including military target detection and advanced predictive maintenance systems, while addressing challenges like the vanishing gradient problem in recurrent neural networks. The use of Long Short-Term Memory (LSTM) networks is emphasized for managing time-dependent data effectively, with guidance on constructing input vectors for training and prediction.
![Input Node
Input Gate
Output Gate
Forgetting Gate
Output Node
u
y
X
σ
φ
Input
Output
State
Sigmoid [0 ÷1]
Tanh [-1 ÷+1]
⊕
+
π
Sum
Concatenation
Product
σ
σ
σ
φ φu yX⊕ +
π
ππ
Long-ShortTerm Memory LSTM](https://image.slidesharecdn.com/enricobustoaddfor-161004093045/85/AI-Engineering-from-Physical-to-Virtual-Sensors-using-Deep-Learning-Enrico-Busto-Add-for-18-320.jpg)
![Input Node
Input Gate
u
y
X
σ
φ
Input
Output
State
Sigmoid [0 ÷1]
Tanh [-1 ÷+1]
⊕
+
π
Sum
Concatenation
Product
σ
φu X⊕ +π
Long-ShortTerm Memory LSTM](https://image.slidesharecdn.com/enricobustoaddfor-161004093045/85/AI-Engineering-from-Physical-to-Virtual-Sensors-using-Deep-Learning-Enrico-Busto-Add-for-19-320.jpg)
![Forgetting Gate
u
y
X
σ
φ
Input
Output
State
Sigmoid [0 ÷1]
Tanh [-1 ÷+1]
⊕
+
π
Sum
Concatenation
Product
σ
u X⊕ +
π
Long-ShortTerm Memory LSTM](https://image.slidesharecdn.com/enricobustoaddfor-161004093045/85/AI-Engineering-from-Physical-to-Virtual-Sensors-using-Deep-Learning-Enrico-Busto-Add-for-20-320.jpg)
![Output Gate
Output Node
u
y
X
σ
φ
Input
Output
State
Sigmoid [0 ÷1]
Tanh [-1 ÷+1]
⊕
+
π
Sum
Concatenation
Product
σ
φu yX⊕ π
Long-ShortTerm Memory LSTM](https://image.slidesharecdn.com/enricobustoaddfor-161004093045/85/AI-Engineering-from-Physical-to-Virtual-Sensors-using-Deep-Learning-Enrico-Busto-Add-for-21-320.jpg)
AI + Engineering: from Physical to Virtual Sensors using Deep Learning, Enrico Busto - Add-for
- 1. AI + ENGINEERING From Physical toVirtual Sensors using DEEP LEARNING #aiwtb
- 2. We developed a novel approach by adapting Deep Learning algorithms to the classical Engineering applications like signal filtering, sensor fusion and control systems. This gave us a competitive advantage over the existing solutions. AI + ENGINEERING
- 3. This is made possible by what we call COMBINATORIAL INNOVATION We Re-Engineer the Deep Learning Algorithms developed for the web and adapt it to work on generic sensor data.
- 4. HUGE demand for Deep LearningVertical Applications in Industrial Environments Examples - some of our projects:
- 5. TIRE DOT CODE RECOGNITION DOT XB 4J R523 4213
- 6. Target Detected Class: Missile Launcher Type: MAZ-543 Uragan MILITARYTARGETS DETECTION AND RECOGNITION
- 7. ADVANCEDTRACTION CONTROLS VEHICLE DYNAMICS DRIVING STYLE DETECTION
- 9. ADVANCED NONLINEAR CONTROLS WASHING MACHINES
- 10. ADAPTIVE LIGHTING CONTROLS EARLY FAULT DETECTION
- 13. Lets focus on FORECASTING
- 14. t y t0 future Typical approach with FFNN
- 16. Recurrent Neural Networks RNN u(t) X y(t) Mind This LOOP !!
- 17. Xu(t0) y(t0) Xu(t1) y(t1) Xu(t2) y(t2) Xu(tn) y(tn) Xu(t2) y(t2) Something Happens at a certain moment in time The INTERNAL STATE keeps the event in memory But the MEMORY of the event is soon forgotten The VANISHING GRADIENT PROBLEM TIME FLOW
- 18. Input Node Input Gate Output Gate Forgetting Gate Output Node u y X σ φ Input Output State Sigmoid [0 ÷1] Tanh [-1 ÷+1] ⊕ + π Sum Concatenation Product σ σ σ φ φu yX⊕ + π ππ Long-ShortTerm Memory LSTM
- 19. Input Node Input Gate u y X σ φ Input Output State Sigmoid [0 ÷1] Tanh [-1 ÷+1] ⊕ + π Sum Concatenation Product σ φu X⊕ +π Long-ShortTerm Memory LSTM
- 20. Forgetting Gate u y X σ φ Input Output State Sigmoid [0 ÷1] Tanh [-1 ÷+1] ⊕ + π Sum Concatenation Product σ u X⊕ + π Long-ShortTerm Memory LSTM
- 21. Output Gate Output Node u y X σ φ Input Output State Sigmoid [0 ÷1] Tanh [-1 ÷+1] ⊕ + π Sum Concatenation Product σ φu yX⊕ π Long-ShortTerm Memory LSTM
- 22. How to build the InputVector for the LSTM
- 23. t y PRESENT FUTUREPAST This is a CONTROLLED VARIABLE: you know PAST and FUTURE IF NEEDED here you can use a causal digital filter The Target Signal should only be filtered with a NON-CAUSAL digital filter to avoid LAG
- 24. This is the INPUT VECTOR for the LSTM u This is the TARGET VALUE for the LSTM (training) and PREDICTION (inference) y This is the TARGET VALUE for the LSTM (training) and PREDICTION (inference) Future Values of the CONTROLLED VARIABLES can be shifted back to present This is the (vanishing) PAST: you do not have to manage it: the internal states of the LSTM will manage (forget) it for you. PRESENT FUTUREPAST
- 25. Remember to Subscribe to the MACHINE LEARNING ITALY Meetup To be Updated on Examples - Code - Benchmarks meetup.com/it-IT/Machine-Learning-Italy
- 26. Visit add-for.com/training-material To download the code and the sample data Contact Me it.linkedin.com/in/ebusto
- 27. Time for CODING