Time series classification (TSC) is the area of machine learning interested in learning how to assign labels to time series. The last few decades of work in this area have led to significant progress in the accuracy of classifiers, with the state of the art now represented by the HIVE-COTE algorithm. While extremely accurate, HIVE-COTE is infeasible to use in many applications because of its very high training time complexity in O(N^2*T^4) for a dataset with N time series of length T. For example, it takes HIVE-COTE more than 72,000s to learn from a small dataset with N=700 time series of short length T=46. Deep learning, on the other hand, has now received enormous attention because of its high scalability and state-of-the-art accuracy in computer vision and natural language processing tasks. Deep learning for TSC has only very recently started to be explored, with the first few architectures developed over the last 3 years only. The accuracy of deep learning for TSC has been raised to a competitive level, but has not quite reached the level of HIVE-COTE. This is what this paper achieves: outperforming HIVE-COTE's accuracy together with scalability. We take an important step towards finding the AlexNet network for TSC by presenting InceptionTime---an ensemble of deep Convolutional Neural Network (CNN) models, inspired by the Inception-v4 architecture. Our experiments show that InceptionTime slightly outperforms HIVE-COTE with a win/draw/loss on the UCR archive of 40/6/39. Not only is InceptionTime more accurate, but it is much faster: InceptionTime learns from that same dataset with 700 time series in 2,300s but can also learn from a dataset with 8M time series in 13 hours, a quantity of data that is fully out of reach of HIVE-COTE.
Today we have the super-effective technique as Transfer Learning where we can use a pre-trained model by Google AI to classify any image of classified visual objects in the world of computer vision. Transfer learning is a machine learning method which utilizes a pre-trained neural network. Inception-v3 is a pre-trained convolutional neural network model that is 48 layers deep. It is a version of the network already trained on more than a million images from the ImageNet database. It is the third edition of Inception CNN model by Google, originally instigated during the ImageNet Recognition Challenge.
Cosmologists are facing the problem of the analysis of a huge quantity of data when observing the sky. The methods used in cosmology are, for the most of them, relying on astrophysical models, and thus, for the classification, they usually use a machine learning approach in two-steps, which consists in, first, extracting features, and second, using a classifier. In this paper, we are specifically studying the supernovae phenomenon and especially the binary classification "I.a supernovae versus not-I.a supernovae". We present two Convolutional Neural Networks (CNNs) defeating the current state-of-the-art. The first one is adapted to time series and thus to the treatment of supernovae light-curves. The second one is based on a Siamese CNN and is suited to the nature of data, i.e. their sparsity and their weak quantity (small learning database).
Special purpose object detection systems need to be fast, accurate and dedicated to classifying a handful but relevant number of objects. Our aim was to integrate a system which utilizes the Inception's vast heuristically mapped image pre-diction tree along-with a real time system accurate and robust enough to work at various processing powers and give the user enough confidence of identifying and detecting object with a single frame. As this property can be utilized at huge number of places relying on real time detection, it might not be limited to only driving assistance or autonomous driving systems, but are beyond the scope of this project. To come up with a novel dataset, which would have an image tree with enough weights and variety so as to predict the objects being identified with high accuracy and precision, was taken up to set up the softmax layer of Inception, which was earlier weighted by the existing ImageNet dataset. The results were a convincing recognition accuracy and prediction confidence with real time test frames of a video.