VISVESVARYA TECHNOLOGICAL UNIVERSITY

Jnana Sangama, Belgavi-590018

PROJECT PHASE-1 REPORT

On
“SEISMOWATCH ADVANCE GUARD USING MACHINE
LEARNING”

submitted in the partial fulfilment of the requirement for the award degree of

BACHELOR OF ENGINEERING
IN
COMPUTER SCIENCE AND ENGINEERING

Submitted By

NINGAMMA KASHINATH BIRADAR - 1AR20CS007

PREKSHA H S -1AR20CS008
PRERANA– 1AR20CS009
SONAL CHATURVEDY – 1AR20CS012

Under the Guidance of

DR. MAHANTESH MATHAPATI,

Professor, CSE Dept.
AIEMS, Bengaluru.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING

2023-24
B.V.V. Sangha’s
AMRUTA INSTITUTE OF ENGINEERING & MANAGEMENT SCIENCES
Bidadi Industrial Area, Bidadi, Bengaluru – 562109
 DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

CERTIFICATE

Certified that the Project phase1 report work entitled “SEISMOWATCH

ADVANCE GUARD USING MACHINE LEARNING” is a bonafide work carried
out by Ningamma Kashinath Biradar,Preksha H.S, Prerana, Sonal Chaturvedy bearing
USN 1AR20CS007, 1AR20CS008, 1AR20CS009, 1AR20CS012 of Amruta Institute of
Engineering and Management Sciences in partial fulfilment for 5th semester in
COMPUTER SCIENCE AND ENGINEERING of the Visveswaraya Technological
University, Belagavi during the academic year 2023-24 It is certified that all
corrections/suggestions Indicated for Internal Assessment have been incorporated in the
report deposited in the departmental library. TheProject phase 1report has been approved as
it satisfies the academic requirements in respect of Project Phase 1 report prescribed by the
institution.

Signature of guide Signature of HOD

DR. MAHANTESH Dr. SRIDHAR
MATHAPATI
Professor and
Professor
Head
Dept. of CSE, AIEMS Dept. of CSE, AIEMS
 ACKNOWLEDGEMENT
The completion of project brings with great sense of satisfaction, but it is never completed
without thanking the persons who all are responsible for its successful completion. First
and foremost, I wish to express my deep sincere feelings of gratitude to our Institution, of
AMRUTA INSTITUTE OF ENGINEERING & MANAGEMENT SCIENCES for
Providing me an opportunity to do my education.

I extend my deep sense of sincere gratitude to Dr. SANTHOSH MURNAL, Principal

AMRUTA INSTITUTE OF ENGINEERING & MANAGEMENT SCIENCES, for having
permitted to carry out the project work successfully.

I express my heartfelt sincere gratitude to Dr. SRIDHAR, Professor & Head of Department
of Computer Science and Engineering, AMRUTA INSTITUTE OF ENGINEERING AND
MANAGEMENT SCIENCES, for his valuable suggestions and support.

I express my special in-depth, heartfelt, sincere gratitude to project guide Prof.

DR. MAHANTESH MATHAPATI, Assistant Professor, Dept. of CS&E, AMRUTA
INSTITUTE OF ENGINEERING & MANAGEMENT SCIENCES, for her constant support.

Finally, I would like to thank all the Teaching Technical faculty and supporting staff members
of Department of Computer Science and Engineering, AMRUTA INSTITUTE OF
ENGINEERING & MANAGEMENT SCIENCES, for their support.

Ningamma Kashinath Biradar

(1AR20CS007)
Preksha H.S
(1AR20CS008)
Prerana
(1AR20CS009)
Sonal Chaturvedy
(1AR20CS012)
 ABSTRACT

Our project aims to improve the accuracy of Earthquake Early Warning systems by means of machine
learning. These are designed to detect and characterize medium and large earthquakes before their
damaging effects reach a certain location. Traditional EEW methods based on seismometers fail to
accurately identify large earthquakes due to their sensitivity to the ground motion velocity. The recently
introduced high-precision GPS stations, on the other hand, are ineffective to identify medium
earthquakes due to its propensity to produce noisy data. In practice, EEW can be seen as a typical
classification problem in the machine learning field: multisensory data are given in input, and
earthquake severity is the classification result. It is distinguished from earthquake prediction in that the
earthquake has nucleated to provide detectable ground motion when and EEW is issued. In this project,
we introduce the Distributed Multi-Sensor Earthquake Early Warning (DMSEEW) system, a novel
machine learning-based approach that combines data from both types of sensors (GPS stations and
seismometers) to detect medium and large earthquakes.
 INDEX

CHAPTER CONTENT PAGE NO

Acknowledgement

Abstraction

01 Introduction 1-2

02 Literature Review 3-4

03 Problem Statement 5-6

04 Objectives 7-9

05 Methodology 10-12

06 Expected Results 13

Conclusion 14

References 15
Seismowatch Advance Guard Using Machine Learning 2023-24

CHAPTER 1
INTRODUCTION

Earthquakes cause substantial loss of life and damage to the built environment across areas
spanning hundreds of kilometers from their origins. These large ground motions often lead to
hazards such as tsunamis, fires and landslides. Earthquake early warning (EEW) is the delivery
of ground shaking alerts or warnings. It is distinguished from prediction in that the earthquake
has nucleated to provide detectable ground motion when an EEW is issued. The warning time
available is the time between detection and when ground motion is experienced by a user.
Potential warning times are therefore seconds to minutes. Likewise, the time available to collect
and process geophysical data and deliver alerts is seconds to minutes, and the actions of users
must be possible in seconds to minutes. To mitigate the disastrous effects, a number of
Earthquake Early Warning (EEW) systems have been built around the world (Allen and Melgar
2019). These critical systems, operating 24/7, are expected to automatically detect and
characterize earthquakes as they happen, and to deliver alerts before the ground motion actually
reaches sensitive areas so that protective measures could be taken. A recent review by (Allen
and Melgar 2019) identified the detection of the whole spectrum of earthquakes with damaging
potential and particularly large earthquakes as an outstanding problem in the field of EEW. An
EEW system needs to be able to detect both medium (5 ≤ magnitude < 6, Richter scale) and
large earthquakes (6 ≤ magnitude, Richter scale). Depending on the distance from the origin of
the earthquake, both of these can cause serious damages. Seismometers, which have long been
the bulwark of seismology to detect earthquakes, have a difficulty to detect and characterize
large earthquakes (Melgar et al. 2013) due to a well-known saturation issue caused by their
sensitivity to ground motion velocity. As a result, earthquakes over magnitude 7.5 tend to be
underestimated. A promising solution to this issue (Melgar et al. 2015) emerged with novel high-
precision Global Positioning System (GPS) sensors, with their millimeter to centimeter
accuracy. Traditionally, inertial seismometers are used to detect primary waves. The inertial
mass is designed to remain stationary following sudden movements while the frame and drum
move with the ground to record waves.

Dept of CSE AIEMS, Bangalore Page 1

Seismowatch Advance Guard Using Machine Learning 2023-24

Rapid growth in the quantity of seismic data has posed challenges for processing and analysis
techniques in modern seismology. Many popular techniques used in major data centres and
seismological research originated from the age when seismic data were small and computational
power was limited. Today machine learning provides a large collection of tools to extract valuable
information from voluminous data. Trained with sufficient data, machine learning algorithms have
human-like capability to recognize natural objects and make expert-level decisions in various
disciplines. In addition, although the computational cost for training is usually high, their online
operation is low cost (e.g., Perol et al., 2018). These advantages make machine learning particularly
suitable for applications in real-time seismology and earthquake early warning.

Dept of CSE AIEMS, Bangalore Page 2

Seismowatch Advance Guard Using Machine Learning 2023-24

CHAPTER 2
LITERATURE SURVEY

➢ A Distributed Multi-Sensor machine Learning approach to earthquake Early detection:

• Author Name: Kevin Fauvel, Diego Melgar and Pedro Silva
• Contribution: To improve the accuracy of EEW systems by means of machine learning. It is
designed to detect and characterize medium and large earthquakes before their damaging
effects reach a certain location. Traditional EEW methods based on seismometers fail to
accurately identify large earthquakes due to their sensitivity to the ground motion velocity. The
recently introduced high-precision GPS stations, on the other hand, are ineffective to identify
medium earthquakes due to its propensity to produce noisy data. In this paper, we introduce
the Distributed Multi-Sensor Earthquake Early Warning (DMSEEW) system, a novel machine
learning-based approach that combines data from both types of sensors (GPS stations and
seismometers) to detect medium and large earthquakes. the combined-sensors (GPS and
seismometers) approach that adopts the rule of relative strength.
• Limitation: It requires advance technology. The combination of two sensors can be difficult to
implement.

➢ Earthquake detection &location for Earthquake Early Warning Using Deep

Learning:
• Author Name: S. Anggraini, S.K. Wijaya1, Daryono3
• Contribution: Earthquake Early Warning System (EEWS) is a warning system that provides
information about the estimated S wave arrival time, which can cause significant and
destructive seismic energy using the information carried by the P wave. Technological
advances in analyzing data supported by big data, the interconnection between networks, and
high-performance computing systems in the era of the 4.0 industrial revolution have posed
challenges to process and analyze earthquake early warning using modern seismological
techniques. we implement deep learning for early detection and classification of the
earthquake P wave and noise signals using raw historical data from 3 component BMKG single
station
• Limitation: It requires large amounts of labeled data for training the system.

➢ Machine Learning Seismic Wave Discrimination: Application to Earthquake Early

detection:
• Author Name: Ze Feng Li, Men-Andri Meier, Egill Hauksson, Zhong Wen Zhan, Jennifer
Andrews

Dept of CSE AIEMS, Bangalore Page 3

Seismowatch Advance Guard Using Machine Learning 2023-24

• Contribution: We have constructed an earthquake P wave discriminator to address the noise

trigger problem in EEW, by making use of two powerful machine learning algorithms, GANs
and Random Forests. The GAN is trained to learn the primary features of early P waves, which
saves us from having manually to define and select the waveform features. Random Forests
take advantage of these features and classify the waveform types with high accuracy. Both
algorithms were trained with a large amount of seismic data, that is, 300,000 P waves and more
than 350,000 noise waveforms recorded in southern California and Japan. Combining the GAN
critic and the Random Forest, we achieved the start-of-the-art performance in discriminating
earthquakes against other impulsive noise triggers, which can significantly reduce false triggers
in EEW. Our study makes a compelling case that GANs have capability to discover compact
representation of seismic waves, which has potential for wide applications in seismology.
• Limitation: It heavily rely on the quality and quantity of the training data. If the data is biased,
incomplete, the model's performance can be compromised.

➢ Earthquake Early Warning: Advances, Scientific challenges, and Societal Needs:

• Author Name: Richard M. Allen and Diego Melgar
• Contribution: EEW methodologies fall into three categories, point source, finite fault, and
ground motion models, and we review the advantages of each of these approaches. The
algorithms in use fall into four categories. Point source algorithms are in use in all regional
systems and provide earthquake information in a few seconds. In larger earthquakes (M > 7),
there is time for finite source algorithms to provide unsaturated magnitude estimates and
describe the fault geometry, both of which lead to more accurate alerts. GMMs provide the most
accurate estimate of forthcoming shaking by forward predicting observed shaking, but they are
limited to a few seconds of warning. Finally, onsite approaches are the simplest in that they use
a sensor at one location to warn the same location, but they have limited accuracy and warning
time.
• Limitations: There are limits to the accuracy of estimating seismic intensity by statistical
attenuation formula, as well as limits to the prediction of land surface amplification.

Dept of CSE AIEMS, Bangalore Page 4

Seismowatch Advance Guard Using Machine Learning 2023-24

CHAPTER 3

PROBLEM STATEMENT

➢ A recent review by (Allen and Melgar 2019) identified the detection of the whole spectrum
of earthquakes with damaging potential and particularly large earthquakes as an
outstanding problem in the field of EEW.

➢ An EEW system needs to be able to detect both medium (5 ≤ magnitude < 6, Richter scale)
and large earthquakes (6 ≤ magnitude, Richter scale).

➢ Depending on the distance from the origin of the earthquake, both of these can cause
serious damages. Seismometers, which have long been the bulwark of seismology to detect
earthquakes, have a difficulty to detect and characterize large earthquakes (Melgar et al.
2013) due to a well-known saturation issue caused by their sensitivity to ground motion
velocity. As a result, earthquakes over magnitude 7.5 tend to be underestimated.

➢ A promising solution to this issue (Melgar et al. 2015) emerged with novel high- precision
Global Positioning System (GPS) sensors, with their millimeter to centimeter accuracy
when measuring high ground motion velocity.

➢ However, GPS are unable to characterize medium earthquakes, as they are prone to
containing significant signals from a variety of noise sources, mostly of atmospheric origin.

➢ Consequently, multisensory solutions (leveraging both GPS and seismometers) appear as

a promising approach. As EEW can be assimilated as a classification problem, where the
input is sensor data and the output is a class (normal activity/medium earthquake/large
earthquake); recent machine learning approaches designed to combine large volumes of
data from multiple data sources can be applied.

➢ Integrating and processing high-frequency data streams from multiple sensors scattered
over a large territory in a timely manner requires high-performance computing techniques
and equipment’s.

➢ EEW systems are designed to detect and characterize medium and large earthquakes before
their damaging effects reach a certain location. Traditional EEW methods based

Dept of CSE AIEMS, Bangalore Page 5

Seismowatch Advance Guard Using Machine Learning 2023-24

on seismometers fail to accurately identify large earthquakes due to their sensitivity to the

ground motion velocity.

➢ The recently introduced high-precision GPS stations, on the other hand, are ineffective to
identify medium earthquakes due toits propensity to produce noisy data.
➢ GPS stations and seismometers may be deployed in large numbers across different locations
and may produce a significant volume of data consequently, affecting the response time and
the robustness of EEW systems.
➢ EEW can be seen as a typical classification problem in the machine learning field: multi-sensor
data are given in input, and earthquake severity is the classification result.
➢ Distributed Multi-Sensor Earthquake Early Warning (DMSEEW) system, a novel machine
learning-based approach that combines data from both types of sensors (GPS stations and
seismometers) to detect medium and large earthquakes.
➢ Our experiments show that DMSEEW is more accurate than the traditional seismometer only
approach and the combined-sensors (GPS and seismometers) approach that adopts the rule of
relative strength.

Dept of CSE AIEMS, Bangalore Page 6

Seismowatch Advance Guard Using Machine Learning 2023-24

CHAPTER 4

OBJECTIVES

The objectives of Seismowatch advance guard are as follows:

➢ P-Wave Detection:
• Develop a system capable of accurately detecting P-waves in seismic data. Implement
signal processing techniques to enhance the precision of P-wave identification.

• Seismic Sensors: Networks of seismic sensors are strategically placed in regions prone
to earthquakes. These sensors detect ground motion caused by seismic waves.

• P-wave Identification: When an earthquake occurs, the sensors detect the initial P-
waves, which are characterized by their compressional motion and higher frequency
compared to other seismic waves.
➢ Data Processing:
• Establish a robust data processing pipeline to handle seismic information efficiently.
Clean and preprocess seismic data to ensure high-quality input for subsequent stages.

• Data Collection: The seismic sensors continuously collect and transmit raw data,
capturing the movement of the Earth in multiple directions. This data is time-stamped
and sent to centralized servers for analysis.

• Real-time Data Transmission: The data from multiple sensors are collected in real-time
and sent to a central processing system, ensuring swift access to information from
various locations.

• Feature Extraction: Various features of the seismic waves are extracted from the
preprocessed data, such as wave amplitude, frequency content, and arrival times of
different wave types (like P-waves and S-waves).
➢ Database Creation:
• Create comprehensive databases for training, testing, and validating neural networks.
Organize the datasets to reflect a diverse and representative range of seismic
conditions.

• Indexing and Organization: The database is organized and indexed based on various
parameters like time, location, magnitude, and waveform characteristics.

Dept of CSE AIEMS, Bangalore Page 7

Seismowatch Advance Guard Using Machine Learning 2023-24

➢ Neural Network Training:

• Utilize state-of-the-art neural network architectures for earthquake prediction. Train the
neural networks using the created databases to recognize patterns with earthquake
precursor signals.
➢ Threshold Comparison:
• Define a threshold value for seismic activity that serves as a benchmark for earthquake
prediction. Implement a comparison mechanism to assess whether the detected seismic
patterns surpass the defined threshold.

• Setting Thresholds: Thresholds are established based on the characteristics of the recorded
data. These thresholds determine the minimum level of ground motion required to trigger
an alert. They are set by analyzing the amplitude, frequency, and duration of seismic waves
during typical non-seismic activities.

• Threshold Adjustment: Periodic review and adjustment of thresholds are necessary to

adapt to changing environmental conditions or to improve the system's sensitivity without
raising false alarm rates. Machine learning algorithms may also help dynamically adjust
thresholds based on ongoing data patterns.

• Event Confirmation: Once seismic activity surpasses the established thresholds and passes
initial filtering, the system proceeds with further analysis to confirm the event's nature,
location, depth, and magnitude.
➢ Prediction and Alert System:
• Integrate the neural network model with a real-time prediction system. Develop an alert
mechanism that triggers when seismic activity exceeds the established threshold.
➢ Communication with Authorities:
• Implement a communication protocol to send alerts promptly to the relevant
authorities. Ensure seamless integration with existing alert systems for efficient
dissemination of information.
• Alert Generation: When seismic activity is detected the early detection system generate
alerts based on the analysis of seismic data, estimating the earthquake's location,
magnitude, and potential impact.

Dept of CSE AIEMS, Bangalore Page 8

Seismowatch Advance Guard Using Machine Learning 2023-24

• Designated Alert Recipients The system is programmed to notify designated authorities

responsible for emergency management, disaster response, public safety agencies,
government bodies, and relevant stakeholders.

• Notification Protocols: Alerts are transmitted via various communication channels such as
dedicated emergency alert systems, specialized software, emails, SMS, phone calls, mobile
apps, sirens, radio, television broadcasts, and social media platforms.

Dept of CSE AIEMS, Bangalore Page 9

Seismowatch Advance Guard Using Machine Learning 2023-24

CHAPTER 5

METHODOLOGY

➢ This project involves the creation of seismic sensor which is the combination of ADXL
senser and Global Positioning System (GPS).
➢ Using these two combinations we are detecting P-wave signals.
➢ The GPs and Seismic sensor are used for data collection.
➢ The ADXL series of sensors, developed by Analog Devices, are accelerometers that measure
acceleration in one, two, or three axes.
➢ These sensors are commonly used in various applications to detect changes in acceleration,
tilt, and motion.
➢ Global Positioning System (GPS) technology has become ubiquitous in various applications,
providing accurate positioning and navigation information.
➢ The data can be processed using Microcontrollers and Microcontroller are versatile
integrated circuits that are designed to execute specific tasks in embedded systems. They
find application in a wide range of electronic devices due to their ability to process data,
control operations, and interact with other components.
➢ Data storage is used to collect real time data & train the system.
➢ Here we are using a CNN. The CNN (Convolutional Neural Networks) are a type of deep
learning algorithm specifically designed for image recognition and processing.
➢ "Alert" and "Broadcast" are critical components for notifying the public and relevant
authorities about potential seismic threats.

ADXL
Sensor Data
Micro Storage CNN Alert
controller Broad
GPS casting

Seismic Sensor
Fig 5.1: Block Diagra

Dept of CSE AIEMS, Bangalore Page 10

Seismowatch Advance Guard Using Machine Learning 2023-24

ADXL Sensor: The ADXL sensors can contribute to earthquake monitoring and early warning
systems, it's important to note that specialized seismometers are typically more precise for detecting
seismic events. These sensors are designed specifically for earthquake monitoring and can provide
more accurate data for seismic analysis. Nonetheless, accelerometers like the ADXL series can still
play a role in earthquake monitoring systems, especially when integrated into larger sensor network

GPS: GPS stations are installed in seismically active regions to measure movements in the Earth's crust.
These stations continuously monitor ground displacement, which can be indicative of tectonic plate
movements or strain accumulation along faults.

Micro controller: Microcontrollers interface with various sensors like accelerometers, seismometers,
GPS modules, or other seismic sensors. They collect raw data from these sensors that detect ground
motion, seismic waves, or other earthquake-related parameters. Microcontrollers compare the observed
seismic activity against predefined thresholds. When the activity exceeds these thresholds, it can trigger
further analysis or initiate alerts.

Data Storage: To ensure data integrity and availability, redundant storage and backup systems are
often employed. Multiple copies of the seismic data might be stored in different physical locations or
on separate servers to prevent data loss due to system failures or disasters.

CNN: While direct earthquake prediction remains challenging, CNNs might be involved in statistical
modeling using historical seismic data. They could assist in identifying trends or patterns that could
contribute to probabilistic seismic hazard assessments.

Alert Broad casting: Alert broadcasting in earthquake early warning systems aims to provide timely
information to individuals and authorities, allowing for quick responses that can mitigate risks and
reduce the impact of seismic events on lives and infrastructure.

Dept of CSE AIEMS, Bangalore Page 11

Seismowatch Advance Guard Using Machine Learning 2023-24

SYSTEM REQUIREMENTS

➢ HARDWARE REQUIREMENTS

• ADXL Sensor

• Seismic Sensor

• GPS (Global Positioning System)

• Micro controller

• Data Storage

• Alert Broad casting

➢ SOFTWARE REQUIREMENTS

• A Distributed Machine Learning Concept.

• Distributed Multi-Sensor Earthquake Early Warning algorithm (DMSEEW).

• CNN (Convolutional neural network)

Dept of CSE AIEMS, Bangalore Page 12

Seismowatch Advance Guard Using Machine Learning 2023-24

CHAPTER 6

EXPECTED RESULT

The expected result of implementing the outlined steps for P-Wave detection and earthquake
prediction is the development of a robust and effective earthquake early warning system. The
system should be capable of accurately detecting P-waves, a type of seismic wave that typically
precedes the more damaging S-waves and surface waves associated with earthquakes. Here's a
breakdown of the expected outcomes:

• Accurate identification of P-waves in seismic data through the implementation of

advanced signal processing techniques, leading to improved precision in earthquake
precursor signal recognition.
• Establishment of a reliable data processing pipeline to efficiently handle seismic
information, ensuring that the input data for subsequent stages is of high quality and
suitable for analysis.
• Utilization of state-of-the-art neural network architectures for earthquake prediction,
with the training of the model using the created databases to recognize and learn patterns
indicative of seismic activity.
• Integration of the trained neural network model with a real-time prediction system,
incorporating an alert mechanism that triggers when seismic activity exceeds the defined
threshold, facilitating timely warnings.
• Implementation of a communication protocol to promptly send alerts to relevant
authorities, ensuring seamless integration with existing alert systems for efficient
dissemination of information to aid in disaster preparedness and response.
• Rigorous testing and evaluation of the system's performance, assessing accuracy,
sensitivity, and specificity in earthquake predictions against known seismic events to
validate the effectiveness of the early warning system.
• Provision of training sessions for end-users and relevant personnel on system usage,
establishment of a support system to address issues or inquiries related to the earthquake
prediction system

Dept of CSE AIEMS, Bangalore Page 13

Seismowatch Advance Guard Using Machine Learning 2023-24

CONCLUSION

The use of machine learning methods in seismology is still in an early phase. One area of
development where it demonstrated promising results is earthquake early warning (EEW), i.e.
the characterization of an earthquake before it reaches sensitive areas. Current state-of-the-art
methods based on seismometers data only demonstrated an applicability limited to medium
earthquakes. In contrast, GPS based methods are only suitable for large earthquake detection.
We propose DMSEEW, a novel stacking ensemble approach for characterizing the whole
spectrum of earthquakes with damaging potential by combining both GPS and seismometer
data. Our evaluation on a real-world dataset collected with domain experts demonstrates that
the proposed distributed stacking ensemble approach improves the detection of both medium
and large earthquakes compared to traditional seismometer only approach and the combined
sensors (GPS and seismometers) baseline approach that adopts the rule of relative strength (F1
score: +7% and +6% on medium earthquakes, +45% and +27% on large earthquakes). In
addition, DMSEEW detects all large earthquakes with a precision of 100%. While existing
solutions rely on fully centralized processing of the sensor data, our approach assumes
distributed data processing based on a geographically distributed cyberinfrastructure. This
design significantly reduces the volume of data transmitted in the network, meets the real-time
requirements while increasing reliability of the EEW system.

Dept of CSE AIEMS, Bangalore Page 14

Seismowatch Advance Guard Using Machine Learning 2023-24

REFERENCES

➢ Allen, R. M., and Melgar, D. 2019. Earthquake Early Warning: Advances, Scientific
Challenges, and Societal Needs. Annual Review of Earth and Planetary Sciences 47:361–
388
➢ Baydogan, M. G., and Runger, G. 2014. Learning a Symbolic Representation for
Multivariate Time Series Classification. Data Mining and Knowledge Discovery 29:400–
422.
➢ Bergstra, J.; Yamins, D.; and Cox, D. 2013. Making a Science of Model Search:
Hyperparameter Optimization in Hundreds of Dimensions for Vision Architectures. In
Proceedings of the 30th International Conference on Machine Learning
➢ Bergstra, J.; Bardenet, R.; Bengio, Y.; and Kegl, B. 2011. ´ Algorithms for Hyper-
Parameter Optimization. In Proceedings of the 24th International Conference on Neural
Information Processing Systems.

Dept of CSE AIEMS, Bangalore Page 15