FocusNet: Classifying better by focusing on confusing classes
نویسندگان
چکیده
Nowadays, most classification networks use one-hot encoding to represent categorical data because of its simplicity. However, may affect the generalization ability as it neglects inter-class correlations. We observe that, even when a neural network trained with labels produces incorrect predictions, still pays attention target image region and reveals which classes confuse network. Inspired by this observation, we propose confusion-focusing mechanism address class-confusion issue. Our is implemented two-branch architecture. Its baseline branch generates confusing classes, FocusNet branch, whose architecture flexible, discriminates correct from these classes. also introduce novel focus-picking loss function improve accuracy encouraging focus on The experimental results validate that our effective for common datasets, can benefit current in improving their accuracy.
منابع مشابه
Classifying Isomorphic Residue Classes
We report on a case study on combining proof planning with computer algebra systems. We construct proofs for basic algebraic properties of residue classes as well as for isomorphisms between residue classes using di erent proving techniques, which are implemented as strategies in a multi-strategy proof planner. We show how these techniques help to successfully derive proofs in our domain and ex...
متن کاملClassifying Problems into Complexity Classes
A fundamental problem in computer science is, stated informally: Given a problem, how hard is it?. We measure hardness by looking at the following question: Given a set A whats is the fastest algorithm to determine if “x ∈ A?” We measure the speed of an algorithm by how long it takes to run on inputs of length n, as a function of n. For example, sorting a list of length n can be done in roughly...
متن کاملClassifying Arabic Verbs Using Sibling Classes
In the effort of building a verb lexicon classifying the most used verbs in Arabic and providing information about their syntax and semantics (Mousser, 2010), the problem of classes over-generation arises because of the overt morphology of Arabic, which codes not only agreement and inflection relations but also semantic information related to thematic arity or other semantic information like ”i...
متن کاملMotivic classes of some classifying stacks
We prove that the class of the classifying stack BPGLn is the multiplicative inverse of the class of the projective linear group PGLn in the Grothendieck ring of stacks K0(Stackk) for n = 2 and n = 3 under mild conditions on the base field k. In particular, although it is known that the multiplicativity relation {T} = {S} · {PGLn} does not hold for all PGLntorsors T → S, it holds for the univer...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Pattern Recognition
سال: 2022
ISSN: ['1873-5142', '0031-3203']
DOI: https://doi.org/10.1016/j.patcog.2022.108709